Bailey S.Simmons,Jessica M.Wempen

Department of Mining Engineering,University of Utah,UT 84112,United States

Keywords:DInSAR Subsidence Monitoring Longwall Trona

ABSTRACT Surface subsidence that results from longwall mining can be large magnitude and can affect significant areas.Conventional methods for subsidence monitoring include leveling,global positioning system(GPS),and photogrammetric surveys.Remote sensing techniques including,aerial LiDAR,terrestrial laser scanning,and satellite-based Differential Interferometric Synthetic Aperture Radar (DInSAR),are also used to measure deformation associated with subsidence.DInSAR data are different than data from conventional subsidence surveys.Images capture data over large areas (hundreds of kilometers),and each pixel (data point) in an image quantifies the average displacement over an area of square meters.DInSAR data can have fairly high time resolution;imaging periods typically range from weeks to months.DInSAR data can be useful to monitor subsidence sequentially over short periods.Regularly monitoring subsidence may help define if caving is progressing normally and can establish relationships between surface deformation and longwall face advance,which has potential to help quantify possible risks to mine stability.In this study,subsidence at a longwall trona mine is monitored over short periods,typically 12 days,as the longwall face is advanced through a panel.C-band interferometric wide swath synthetic aperture radar (SAR) images from the sentinel satellites are used to quantify the subsidence.The onset of subsidence occurs close in time to the beginning of the longwall face advance,and overall,the development of subsidence closely follows the longwall face advance.

1.Introduction

Longwall mining is a caving method.As a panel is mined out,the immediate roof caves forming the gob.Strata failure propagates to the surface almost immediately causing subsidence [1].When subsidence is planned as part of longwall mine development,monitoring surface deformation can be an important part of a subsidence control plan [2,3].

Conventional surveying methods are extremely accurate.GPS surveys typically have horizontal and vertical accuracies of 5 and 20 mm,respectively.Leveling surveys are very accurate over small areas,with a vertical resolution from 0.1 to 1 mm[4].Surface models developed using photogrammetry with images collected using unmanned aerial vehicles(UAVs)can have accuracies of 3 cm horizontally and 5 cm vertically,under ideal conditions [5].

Differential interferometric synthetic aperture radar(DInSAR)is a satellite-based remote sensing technique that can be used to measure surface displacement on the Earth.Applying DInSAR,phase measurements from radar images acquired over the same region at different times are used to precisely measure relative distances[6].DInSAR is sensitive to very small changes and can measure displacement with accuracies ranging from millimeters to centimeters under ideal conditions [7,8].Some conditions that can contribute to high accuracy DInSAR include moderate topography,stable atmospheric conditions,regular surface conditions with limited changes in vegetation and ground cover,and relatively slow rates of deformation.Subsidence monitoring is one application of DInSAR.Often,for precise measurement of subsidence using DInSAR,conditions are not ideal,particularly when large magnitude subsidence occurs quickly.Interferometric data processing methods and data accessibility continue to improve,and DInSAR continues to be evaluated as a method to quantify subsidence related to active and historic mining.Recent studies by Wempen et al.focus on applying DInSAR for subsidence monitoring [9-12].

DInSAR has the potential to supplement conventional subsidence surveys.Using DInSAR monitoring can be carried out regularly over short period,and the data have high spatial resolutions.For satellite-based SAR data,high resolution data often have pixel areas on the ground surface that are hundreds of square meters;the sentinel data used in this study have a ground (pixel)resolution of 5 m by 20 m(100 m2).DInSAR has potential to monitor the magnitude,extent,and rate of subsidence over short periods,and also to identify periods when subsidence is minimal.Applying DInSAR,deformation is measured by pixel,and each measurement represents the average deformation of the pixel.Horizontal accuracy is sub-pixel.DInSAR does have limitations for subsidence monitoring.Changing surface conditions,including variable snow cover and changes in vegetation,as well as rugged topography can negatively affect the quality of DInSAR data.Additionally,large displacements can lead to phase saturation,which limits the interpretability of the displacement data [13].

In this study,subsidence at a longwall trona mine is monitored over short periods,typically 12 days,as the longwall face is advanced through a panel.Monitoring was carried out for about six months.Longwall face locations were reported quarterly.The objectives of this study are to use DInSAR to quantify the development of subsidence in short periods and to compare the development of subsidence to longwall face advance.

2.Study region:Southwest Wyoming

Trona is an important evaporate mineral used to produce soda ash.The Green River Basin in southwest Wyoming contains one of the largest global resources of natural soda ash [14].The trona deposits in the basin are extensive and shallowly dipping,and mining has historically been carried out using longwall and room and pillar methods at depths from 245 to 520 m [15,16].

There are currently two longwall trona mines operating in the Green River Basin.In 2015,longwall mining produced about 8.1 million tonnes of trona[17].Fig.1 shows an overview of the longwall panel monitored as part of this study.Longwall face locations on June 28,2017,September 30,2017 and January 1,2018,are noted.Using the DInSAR data,subsidence profiles were generated along section A-A’.The panel is approximately 200 m wide and 2690 m long.The mining height is about 3 m.Depth of the panel is approximately 480 m.In this region,the maximum subsidence reported by the mine is about 1.5 m.

3.Data and processing

The C-band (6 cm) data used in this study were imaged by the sentinel 1A satellite.Sentinel 1A was launched in 2014,and is operated by the European Space Agency (ESA).Images were collected in interferometric wide mode.Images have a swath width of 250 km and a pixel resolution of 5 m by 20 m.All of the data used in this study are from orbit 27 and path 454.The orbit is descending.The image dates used in this study are summarized in Table 1.

SAR data processing was performed using the software SARscape®and ENVI®(SARscape®2014).First,single look complex(SLC)image data acquired from vertex[18]and orbit files acquired from the European Space Agency were imported into SARscape[19].Next,paired images were co-registered and differential interferograms were formed.The differential interferograms contain displacement data as phases.In the differential interferograms the phases due to topography and changes in the satellite orbits have been removed.Next,the differential interferograms were filtered to reduce noise.The filtered images were then unwrapping to convert the phases into relative displacements.Finally,the images were geocoded,to convert from radar coordinates to real coordinates,and the displacements were evaluated.Importantly,using DInSAR,displacement is estimated along the satellite line of sight,and to estimate subsidence,motion is assumed to be vertical.

Table 1.entinel 1A data acquisition dates.

4.Results

Sections along A-A’ are shown in Fig.2a,b,and c.Fig.2a illustrates a profile of subsidence from June 5 to June 29;Fig.2b illustrates profiles of subsidence over 12 day intervals from June 29 to October 3;and Fig.2c illustrates profiles of subsidence over 12 day intervals from October 3,2017 through January 7,2018.Dashed vertical lines in Fig.2 illustrate the longwall face locations along section A-A’(Fig.1).Phase saturation occurred in all of the differential interferograms from August 16,2017,through January 7,2018.During the 12 day monitoring periods between these dates,in areas,subsidence developed too rapidly to be measured precisely using DInSAR.Areas affected by phase saturation were clipped from the final images.

Subsidence occurs almost immediately as the longwall face begins to advance through the panel;initially the vertical displacement is small.During the monitoring period from June 5 to June 29(24 day),the maximum vertical displacement is about 2 cm(Fig.2a).After July 11,the magnitude of the vertical displacement increases,and the subsidence begins to develop in the direction of the of the longwall face advance.During the monitoring period from August 4 through August 16,the maximum vertical displacement is about 9 cm (Fig.2b).For the periods from August 16 through January 7 (Fig.2b and c),higher rates of subsidence,greater than 10 cm in 12 day,cause phase saturation in the DInSAR data.In these images,the boundaries of subsidence are defined,but the maximum magnitudes cannot be precisely measured.

Fig.1.Longwall panel layout.Longwall face locations on June 28,2017,September 30,2017,and January 1,2018 are noted.

Fig.2.Profiles of subsidence along section A-A’ for image periods from June 5 to June 29,2017,June 29 to October 3,2017,and October 3,2017 to January 7,2018.

Fig.3 summarizes the longwall face advance rate and the linear rate of subsidence progression.In Fig.3,day zero represents June 5,2017.The subsidence progression is defined based on the distance between consecutive points of maximum deformation in sequential DInSAR images.The distances were estimated in one dimension,the direction of longwall face advance.The maximum subsidence magnitude cannot be measured precisely in the data affected by phase saturation,but the approximate location of the maximum point of deformation was estimated.To estimate the point of maximum deformation,profiles of subsidence were created along the long axis of the longwall panel.On each side of the subsidence profiles,lines were drawn parallel and adjacent to the slopes of the subsidence profiles.The points where the lines intersect provide estimates of the locations of maximum deformation.Estimated data are represented in green in Fig.3.

For the period from June 28 to September 30,2017,the longwall face advanced through the panel at an average rate of 5.5 m/d.The average linear rate of subsidence progression over the June to September period is about 4 m/d.For the period from September 30,2017 to January 1,2018 the longwall face advanced through the panel at a rate of about 6.9 m/d.The average linear rate of subsidence progression over this period is about 6.6 m/d.

5.Discussion

The overall conditions in southwest Wyoming are good for application of DInSAR.In the mining region,the topography is consistent and there is limited vegetation.In winter months,particularly from December through February,the quality of the data can be negatively impacted by changing surface conditions due to snow.The subsidence in the region is fairly large magnitude,so even when monitoring is carried over short periods,phase saturation can limit precisely measure subsidence.

In this study,surface deformation above the longwall panel can be detected soon after the longwall face begins to advance through the panel (Fig.2a).As mining progresses,the rate of downward movement increases significantly (Fig.2a,b,and c).While subsidence is measureable at the surface almost immediately,it takes longer,more than 70 days,for the trough to more fully develop.During this period,the linear subsidence progression lags behind the longwall face advance.Over short periods,the linear rate of subsidence progression is variable,but once the subsidence trough is more developed,the average linear rate of subsidence progression is very close to the average longwall face advance rate.

Fig.3.Longwall face advance rate in comparison to linear subsidence advance rate.

6.Conclusions

For small magnitude subsidence,DInSAR can accurately measure surface displacements.Large deformation gradients can make the radar images incoherent,and as the rate of vertical deformation increases,DInSAR becomes less effective for accurately measuring surface displacements.Although,subsidence magnitudes may not be measured precisely,subsidence boundaries can typically be defined.Sentinel 1A SAR data collected using a C-band(6 cm wavelength) sensor,with 5 m by 20 m pixel resolution,and a 12 d imaging period were used in this study.Using these data to monitor subsidence in the Wyoming region,deformation that occurs at a rate of more than about 10 cm per 12 d cannot be precisely quantified.

Although the sentinel 1A data have some limitations for subsidence monitoring,the data are publicly available.Synthetic aperture radar (SAR) data with a longer band,a higher spatial resolution (smaller pixels),or a more frequent imaging period,could be more effective for precisely quantifying subsidence in this mining region.SAR data from other satellites are potentially available,however,data from these satellites are not currently public.

Acknowledgements

Funding for this research was provided by the National Institute for Occupational Health and Safety (NIOSH).The conclusions of this work are those of the authors and do not represent the opinions or policies of NIOSH.SAR images used in the study are from Copernicus Sentinel data from 2017.Raw data were processed by the European Space Agency(ESA);SAR images were retrieved from ASF DAAC in 2019.The contributions of these organizations are gratefully acknowledged.