sensors

Article A Decade of Ground Deformation in Kunming (China) Revealed by Multi-Temporal Synthetic Aperture Radar Interferometry (InSAR) Technique

Wu Zhu 1,* , Wen-Liang Li 1, Qin Zhang 1, Yi Yang 2, Yan Zhang 1, Wei Qu 1 and Chi-Sheng Wang 3,4 1 College of Geology Engineering and Geomatics, Chang’an University, Xi’an 710064, China; [email protected] (W.-L.L.); [email protected] (Q.Z.); [email protected] (Y.Z.); [email protected] (W.Q.) 2 Yunnan Institute of Surveying and Mapping Engineering, Kunming 650033, China; [email protected] 3 The Guangdong Key Laboratory of Urban Informatics, School of Architecture & Urban Planning, Shenzhen University, Shenzhen 518060, China; [email protected] 4 The Key Laboratory of Urban Land Resources Monitoring and Simulation, Ministry of Land and Resources, Beijing 100812, China * Correspondence: [email protected]; Tel./Fax: +86-29-82339261



Received: 18 July 2019; Accepted: 9 October 2019; Published: 11 October 2019


Abstract: Large-scale urbanization has brought about severe ground subsidence in Kunming (China), threatening the stability of urban infrastructure. Mapping of the spatiotemporal variations of ground deformation is urgently needed, along with summarization of the causes of the subsidence over Kunming with the purpose of disaster prevention and mitigation. In this study, for the first time, a multi-temporal interferometric synthetic aperture radar (InSAR) technique with L-band Advanced Land Observation Satellite (ALOS-1) and X-band Constellation of Small Satellites for Mediterranean basin Observation (COSMO-SkyMed) data was applied to Kunming to derive the time series deformation from 2007 to 2016. The annual deformation velocity revealed two severe subsiding regions in Kunming, with a maximum subsidence of 35 mm/y. The comparison of the deformation between InSAR and leveling showed root-mean-square error (RMSE) values of about 4.5 mm for the L-band and 3.7 mm for the X-band, indicating that our results were reliable. We also found that the L-band illustrated a larger amount of subsidence than the X-band in the tested regions. This difference was mainly caused by the different synthetic aperture radar (SAR)-acquired times and imaging geometries between the L- and X-band SAR images. The vertical time series deformation over two severe subsiding regions presented an approximate linear variation with time, where the cumulative subsidence reached 209 mm during the period of 2007–2016. In view of relevant analyses, we found that the subsidence in Kunming was the result of soft soil consolidation, building load, and groundwater extraction. Our results may provide scientific evidence regarding the sound management of urban construction to mitigate potential damage to infrastructure and the environment.

Keywords: ground deformation; multi-temporal InSAR; Kunming; urbanization

1. Introduction Kunming, the capital and largest city of Yunnan province in Southwest China, is located at a longitude of 102◦100–103◦400 east and a latitude of 24◦230–26◦220 north. Situated in the central Yunnan–Guizhou plateau in Southwest China, Kunming is surrounded by the Dian Lake to the south and by mountains on the other three sides [1]. As China’s frontier to Southeast Asia and South

Sensors 2019, 19, 4425; doi:10.3390/s19204425 www.mdpi.com/journal/sensors Sensors 2019, 19, x FOR PEER REVIEW 2 of 31 Sensors 2019, 19, 4425 2 of 29 and by mountains on the other three sides [1]. As China’s frontier to Southeast Asia and South Asia, largeAsia,-scale large-scale urban urbanconstruction construction has been has implemented been implemented in Kunming in Kunming over the over past thetwenty past years. twenty During years. thisDuring procedure, this procedure, a great amou a greatnt of amount cultivated of cultivated land and gardens land and were gardens occupied were by occupied high-rise by buildings, high-rise roads,buildings, and roads, other structures and other structures[2]. However, [2]. thisHowever, large-scale this large-scale urbanization urbanization brought about brought a series about of a geologicalseries of geological hazards, hazards, such as such landslides, as landslides, subsidence, subsidence, and and ground ground fissures. fissures. Particularly, Particularly, ground ground subsidencesubsidence became became one one of of the the most most prominent prominent geological geological hazards hazards and and caused caused substantial substantial damage damage to to homes,homes, roads, roads, canals, canals, pipelines, pipelines, and and other other infrastructure infrastructure [ 3].3]. Figure Figure 1 shows some photos of the effects e ffects ofof subsidence subsidence in in Kunming, Kunming, where where more more than than fifty fifty individual individual damage damage reports reports were were observed observed through through filedfiled investigations investigations organized inin JuneJune 20162016 [[4].4]. ToTo prevent prevent and and mitigate mitigate future future disasters, disasters, it it is is urgent urgent to tocharacterize characterize and and monitor monitor the the spatiotemporal spatiotemporal variations variations ofof groundground deformationdeformation and summarize the the cacausesuses of of subsidence subsidence in in Kunming. Kunming.

FigureFigure 1. Ground 1. Ground deformation deformation investigated investigated in in Kunming Kunming. and (a) the the uplift effects of buildingon buildings steps, and (b) ground roads [4]. crack, (c) ground subsidence, (d) wall crack caused by the groud deformation. Early studies initiated a few years ago using ground-based techniques (e.g., leveling) documentedEarly studies the extent initiated and a causes few years of subsidence ago usingground-based in Kunming [5 techniques,6]. According (e.g., to leveling) their observations, documented thethe subsidence extent and causeswas mainly of subsidence focused inin Kunming two regions, [5,6]. where According the subsiding to their observations, rate reached the 30 subsidencemm/y [5]. Consideringwas mainly focusedlocal geological in two regions, conditions where and the human subsiding activities, rate reached it was reported 30 mm/y that [5]. the Considering subsidence local in Kunminggeological was conditions mainly andcaused human by groundwater activities, it was extraction, reported building that the subsidenceload, and natural in Kunming consolidation was mainly of soilcaused [7]. byAlthough groundwater these extraction,observations building provided load, useful and natural information, consolidation they had of low soil [spatial7]. Although resolution, these andobservations therefore, provided more useful detailed information, and comprehensive they had low groundspatial resolution, deformation and therefore, was difficult more detailed to see. Interferometricand comprehensive synthetic ground aperture deformation radar was (InSAR) difficult has to demonstrated see. Interferometric its potential synthetic for aperture high-density radar spatial(InSAR) mapping has demonstrated of ground deformation its potential forassociated high-density with earthquakes spatial mapping [8], volcanoes of ground [9], deformation and other geologicassociated processes. with earthquakes Recent adva [8], volcanoesnced InSAR [9], techniques and other geologic have improved processes. our Recent understanding advanced InSARof the processtechniques of haveground improved deformation, our understanding such as the of deep the process learning of groundapproach deformation, [10], artificial such intelligence as the deep techniquelearning approach [11], optimal [10], artificialphase unwrapping intelligence algorithm technique [[12],11], optimalsignal retrieval phase unwrapping for decorrelating algorithm targets [12], [13],signal four retrieval-dimensional for decorrelating filtering approach targets [ 13 [14],], four-dimensional improved synthetic filtering aperture approach radar [ 14 (SAR)], improved image coregistrationsynthetic aperture algorithm radar [15], (SAR) atmospheric image coregistration delay correction algorithm method [15 [16],], atmospheric and coherent delay point correction selection algorithmmethod [16 [17].], and Additionally, coherent point an increasing selection algorithmamount of [SAR17]. Additionally,satellites have an provided increasing a large amount set of of multiSAR- satellitessensor SAR have images, provided which a large have set been of multi-sensoremployed to SARreconstruct images, the which spatiotemporal have been employedevolution of to groundreconstruct deformation the spatiotemporal [18]. In this evolution field, current of ground research deformation is focused [ 18on]. three In this aspects. field, currentThe first research aspect is is applyingfocused on the three multi aspects.-sensor The InSAR first to aspect derive is applyingthe long- theterm multi-sensor ground deformation, InSAR to derive such theas in long-term Tehran (Iran)ground [19 deformation,], Xi’an (China) such [20], as Baja in Tehran California (Iran) (Mexico) [19], Xi’an [21], (China)Urayasu [ 20(Japan)], Baja [22], California the Italian (Mexico) Peninsula [21], [23],Urayasu Istanbul (Japan) (Turkey) [22], the [24], Italian and PeninsulaSemarang [ 23(Indonesia)], Istanbul [25]. (Turkey) The [sec24],ond and aspect Semarang is resolving (Indonesia) two- [ 25or]. three-dimensional ground motion through the multi-sensor SAR images merging technique, which

Sensors 2019, 19, 4425 3 of 29 Sensors 2019, 19, x FOR PEER REVIEW 3 of 31

Thehas second been aspect used in is resolving an urban two- subsiding or three-dimensional area [26], mining ground subsiding motion through area the [27], multi-sensor and co-seismic SAR imagesdeformation merging area technique, [28]. The whichthird aspect has been is developing used in an the urban advanced subsiding algorithm area [26 to], miningconnect subsiding the multi- areasensor [27 ],SAR and images co-seismic so that deformation the long- areaterm [28 time]. The series third deformation aspect is developing is observed the [29]. advanced However, algorithm there toare connect no bibliographies the multi-sensor available SAR images for ground so that deformationthe long-term analysis time series with deformation advanced is multi observed-temporal [29]. However,InSAR forthere Kunming are noso bibliographiesfar. available for ground deformation analysis with advanced multi-temporalTo better understand InSAR for Kunming the spatiotemporal so far. evolution of ground deformation in Kunming, a multi- temporalTo better InSAR understand technique was the utilized spatiotemporal for the first evolution time in this of study ground to obtain deformation a decade in of Kunming,time series adeformation multi-temporal of Kunming. InSAR technique For this, 20 was L-ban utilizedd Advanced for the firstLand time Observation in this study Satellite to obtain(ALOS a-1 decade) and 40 ofX- timeband series Constellation deformation of Small of Kunming. Satellites For for this, Mediterranean 20 L-band Advanced basin Observation Land Observation (COSMO-SkyMed) Satellite (ALOS-1)images were and acquired 40 X-band in Constellation ascending and of descending Small Satellites orbits, for respectively. Mediterranean The annual basin Observation deformation (COSMO-SkyMed)velocity was estimated images to werecharacterize acquired the in ascendingspatial pattern and descending of subsidence orbits, in Kunming. respectively. Subsequently, The annual deformationcomparisons velocity were conducted was estimated to validate to characterize our results and the spatialthe time pattern series deformation of subsidence was in interpreted Kunming. Subsequently,in detail. The causes comparisons of subsidence were conducted are discussed to validate through our correlation results and anal theysis. time series deformation was interpreted in detail. The causes of subsidence are discussed through correlation analysis. 2. Geological Setting of the Study Area 2. Geological Setting of the Study Area The study area is located at Kunming, which is the capital of Yunnan province in southwest China,The as study shown area in is Figure located 2. at Situated Kunming, in a which fertile is lake the capitalbasin on of Yunnanthe northern province shore in southwestof Lake Dian China, and assurrounded shown in Figure by moun2. Situatedtains to in the a fertile north, lake west, basin and on east, the northernKunming shore is located of Lake at Dian an altitude and surrounded of 1900 m byabove mountains sea level to and the north,at a latitude west, andjust north east, Kunmingof the Tropic is located of Cancer at an [2]. altitude As of of2014, 1900 Kunming m above has sea a levelpopulation and at aof latitude 6.6 million, just north with ofan theurban Tropic population of Cancer of [ 24.5]. Asmillion of 2014, [30]. Kunming Due to the has low a population latitude and of 6.6high million, elevation, with Kunming an urban populationhas one of the of 4.5mildest million climates [30]. Due in China, to the lowcharacterized latitude and by highshort, elevation, cool, dry Kunmingwinters with has onemild of days the mildest and crisp climates nights, in China,and long, characterized warm, humid by short, summers, cool, dry which winters are withstill much mild dayscooler and than crisp the nights, lowlands. and long, warm, humid summers, which are still much cooler than the lowlands.

Legend

COSMO-SkyMed coverage

ALOS-1 coverage

Water area

Leveling points

Main town 10 km

Figure 2. Study area and synthetic aperture radar (SAR) data coverage used in this study; corresponding Figure 2. Study area and synthetic aperture radar (SAR) data coverage used in this study; data are superimposed on the digital elevation model (DEM). corresponding data are superimposed on the digital elevation model (DEM). The study area is a late Cenozoic graben basin controlled by a number of quaternary active The study area is a late Cenozoic graben basin controlled by a number of quaternary active faults, e.g., the Heilongtan-Guandu fault (F150), Sheshan fault (F152), Pudu river fault (F54) and faults, e.g., the Heilongtan-Guandu fault (F150), Sheshan fault (F152), Pudu river fault (F54) and Puji- Hanjia village fault (F55) [4]. Figure 3 shows the geological map and location of faults in Kunming. The quaternary strata with different sediment thicknesses are widespread in the study area, including

Sensors 2019, 19, 4425 4 of 29

Puji-HanjiaSensors 2019 village, 19, x FOR fault PEER (F55 REVIEW)[4]. Figure 3 shows the geological map and location of faults in Kunming.4 of 31 The quaternary strata with different sediment thicknesses are widespread in the study area, including thethe lower, lower, middle, middle, and and upper upper Pleistocene Pleistocene series series as well as well as the as the Holocene Holocene series. series. Considering Considering the the composition,composition, it isit furtheris further divided divided into into clay, clay, muddy muddy clay, clay, and and peaty peaty soils soils [31 ].[31]. These These soft soft soils soils have have manymany unfavorable unfavorable properties properties regarding regarding their their use use in in projects, projects, such such as as high high compressibility, compressibility, rheology, rheology, andand thixotropy thixotropy [1]. [1].

Xiaoxinjie-Xu jia du Xiaojiang Majie-Nanyangjie Wanshoushan

Beigucheng-Liangshan Lvfeng Village-Pan xi Jiu Village Pu du River

Puji-Hanjia Village Lihuahai aiyi-Hengchong

Heilongtan-Guandu Panlong River Sheshan Yiduoyun

Dachun River Machang-Xianjie

FigureFigure 3. 3.Geological Geological map map of of study study area. area. The The black black solid solid line line represents represents quaternary quaternary active active faults. faults [4]. 3. Materials and Methods 3. Materials and Methods 3.1. SAR Datasets 3.1. SAR Datasets A total of 20 L-band ALOS-1 images with stripe mode and 40 X-band COSMO-SkyMed (CSK) imagesA were total collected of 20 L-band to derive ALOS the-1 timeimages series with deformation stripe mode over and Kunming.40 X-band COSMO Table1 and-SkyMed Figure (CSK)2 showimages the parameters were collected and to coverages derive the of time these series SAR deformation images, respectively. over Kunming. The ALOS-1 Table 1 imagesand Figure with 2 show an ascendingthe parameters orbit and and stripe coverages mode were of acquired these SAR between images, January respectively. 2007 and March The ALO 2011.S-1 To images maintain with the an temporalascending continuity, orbit and CSK stripe images mode with were ascending acquired orbit between were acquiredJanuary 2007 between and JuneMarch 2011 2011. and To January maintain 2016.the Basedtemporal on these continuity, two types CSK of images SAR data, with we ascending were able orbit to obtain were the acquired ground deformationbetween June over 2011 the and studyJanuary area 2016. between Based January on these 2007 two and types January of SAR 2016. data, In we order were to able validate to obtain the obtained the ground deformation, deformation additionalover the SAR study images area and between leveling January data were 2007 employed and January in this 2016.study, In including order to 91 validate Sentinel-1A the images obtained anddeformation, 37 leveling pointsadditional (represented SAR images by red and stars leveling in Figure data2). were A light employed detection in andthis rangingstudy, including (LIDAR) 91 digitalSentinel elevation-1A images model and (DEM), 37 leveling which points had a spatial(represented resolution by red of stars 3 m andin Figure centimeter-level 2). A light heightdetection and ranging (LIDAR) digital elevation model (DEM), which had a spatial resolution of 3 m and centimeter-level height precision, was acquired as an external DEM to remove the topographic phase from the differential interferograms [4].

Sensors 2019, 19, 4425 5 of 29 precision, was acquired as an external DEM to remove the topographic phase from the differential interferograms [4].

Table 1. Parameters of synthetic aperture radar (SAR) data in this study.

Orbit Azimuth Incidence Number of No. Satellite Data Period Direction Angle (◦) Angle (◦) SAR Images 1 ALOS-1 Ascending 10 38 20 09/01/2007–07/03/2011 − 2 COSMO-SkyMed Descending 10 29 40 13/06/2011–06/01/2016 3 Sentinel-1A Ascending 12 39 31 23/01/2015–17/02/2017 − 4 Sentinel-1A Descending 11 39 60 25/05/2015–01/09/2019

3.2. Multi-Temporal InSAR Processing Using the collected 20 L-band ALOS-1 and 40 X-band CSK images, time series of deformation was derived from the multi-temporal InSAR technique with GAMMA Remote Sensing and Consulting AG (GAMMA) SAR software. This process included three steps. Firstly, interferometric pairs were generated by setting small temporal and spatial baselines. This depressed noise effects and maintained a high level of coherence [32]. On the basis of the experiment, a spatial baseline below 1500 m and a temporal baseline less than 1000 days were designed for the L-band SAR images to generate the interferograms. For the X-band SAR images, a spatial baseline below 220 m and a temporal baseline less than 200 days were designed to generate the interferograms. A total of 59 interferograms from the L-band (Figure4a) and 88 interferograms from the X-band (Figure4b) were produced for the time series deformation estimation. Secondly, after selection of the point targets, removal of the topographic and orbital phases, filtering of the interferograms, and retrieval of the absolute phase, unwrapped interferograms were produced. The process of point target selection was based on the method in [33]. The topographic and orbital phases were simulated and then removed from the interferometric phase using collected LIDAR DEM data and SAR orbital data. Then, the process of adaptive spectral filtering with a window size of 64 for the L-band and 32 for the X-band was executed to suppress the interferometric noise [34]. Finally, phase unwrapping using the minimum cost flow (MCF) method was carried out to retrieve the absolute phase [35]. During this procedure, a common point target was selected as the reference point to maintain consistency of results from the L- and C-bands. Thirdly, the time series deformation was estimated from the collected L- and X-band images. After the second step, the unwrapped interferometric phase may consist of residual phase errors, such as topographic phase error, orbital error, or atmospheric delay error. Prior to the time series deformation estimation, these residual phase errors should be removed from the unwrapped interferometric phase. In this case, the topographic phase error was safely neglected since high-precision LIDAR DEM was used to simulate the topographic phase. A quadratic polynomial phase model was fitted to reduce the orbital phase error [36] and the residual atmospheric delay error, including atmospheric and ionospheric delay, was mitigated by spatial low-pass filtering and temporal high-pass filtering [33]. After these processes, the residual phase errors were finally reduced and the high-precision time series deformation was produced. Sensors 2019, 19, 4425 6 of 29

(a) baselines from ALOS-1/PALSAR-1

(b) baselines from COSMO-SkyMed

Figure 4. The distribution of the temporal and spatial baselines from (a) Advanced Land Observation SatelliteFigure (ALOS-1) 4. The distribution and (b) Constellation of the temporal of Small and spatial Satellites baselines for Mediterranean from (a) Advanced basin Land Observation Observation (COSMO-SkyMed)Satellite (ALOS in-1) this and study. (b) Constellation of Small Satellites for Mediterranean basin Observation (COSMO-SkyMed) in this study. 4. Results and Analysis

4.1. Annual Deformation Velocity The annual deformation velocity maps in the vertical direction were derived from 59 L-band interferograms and 88 X-band interferograms, as shown in Figure5; the left figure shows the result from the L-band ALOS-1 and the right figure shows the result from the X-band CSK. It should be noted that we assumed that the deformation of Kunming was dominated by the vertical direction and was insignificant in the horizontal direction when converting the line-of-slight (LOS) deformation into the vertical deformation. Actually, it was acceptable for our region as the deformation was mostly vertical [5–7]. The statistics in Figure5 show that a large portion of the deformation measurements was strongly skewed toward negative values, indicating that most areas of Kunming are in a state Sensors 2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors Sensors 2019, 19, x FOR PEER REVIEW 7 of 31

such as topographic phase error, orbital error, or atmospheric delay error. Prior to the time series deformation estimation, these residual phase errors should be removed from the unwrapped interferometric phase. In this case, the topographic phase error was safely neglected since high- precision LIDAR DEM was used to simulate the topographic phase. A quadratic polynomial phase model was fitted to reduce the orbital phase error [36] and the residual atmospheric delay error, including atmospheric and ionospheric delay, was mitigated by spatial low-pass filtering and temporal high-pass filtering [33]. After these processes, the residual phase errors were finally reduced and the high-precision time series deformation was produced.

4. Results and Analysis

4.1. Annual Deformation Velocity The annual deformation velocity maps in the vertical direction were derived from 59 L-band interferograms and 88 X-band interferograms, as shown in Figure 5; the left figure shows the result from the L-band ALOS-1 and the right figure shows the result from the X-band CSK. It should be noted that we assumed that the deformation of Kunming was dominated by the vertical direction and was insignificant in the horizontal direction when converting the line-of-slight (LOS) Sensorsdeformation2019, 19, 4425 into the vertical deformation. Actually, it was acceptable for our region as the7 of 29 deformation was mostly vertical [5‒7]. The statistics in Figure 5 show that a large portion of the deformation measurements was strongly skewed toward negative values, indicating that most areas of subsidence. High subsidence rates were observed to be concentrated in two regions (Xishan and of Kunming are in a state of subsidence. High subsidence rates were observed to be concentrated in Guandu towns, which are marked with black rectangles in Figure5), where the maximum subsidence two regions (Xishan and Guandu towns, which are marked with black rectangles in Figure 5), where reached 35 mm/y using ALOS-1 (left of Figure5) and 30 mm /y using X-band COSMO-SkyMed (right the maximum subsidence reached 35 mm/y using ALOS-1 (left of Figure 5) and 30 mm/y using X- ofband Figure COSMO5). -SkyMed (right of Figure 5).

ALOS-1 Reference point COSMO-SkyMed Reference point Xishan Xishan

Guandu Guandu

10 km 10 km

Figure 5. The annual deformation velocity maps from ALOS-1 (left) and COSMO-SkyMed (right) in this Figure 5. The annual deformation velocity maps from ALOS-1 (left) and COSMO-SkyMed (right) in study. The red cross shows the reference point and the black rectangles denote the subsiding region. this study. The red cross shows the reference point and the black rectangles denote the subsiding region.

CarefulCareful inspection inspection of of Figure Figure5 5suggests suggests that that there there was was a a subtle subtle difference di fference in in the the subsiding subsiding magnitudemagnitude between between the the L- L and- and X-band X-band results; results; the the L-band L-band (left) (left) illustrated illustrated a a larger larger subsidence subsidence than than the the X-band. This phenomenon was particularly obvious for the marked region in Figure 5. This X-band. This phenomenon was particularly obvious for the marked region in Figure5. This di fference difference may have been caused in two ways. The first reason involves the different SAR-acquired may have been caused in two ways. The first reason involves the different SAR-acquired times, i.e., times, i.e., the L-band result was derived between January 2007 and March 2011 while the X-band the L-band result was derived between January 2007 and March 2011 while the X-band result was result was derived between June 2011 and January 2016. Clearly, there was no overlapping time derived between June 2011 and January 2016. Clearly, there was no overlapping time between the L- between the L- and X-bands. As shown in the introduction, the subsidence in Kunming was highly and X-bands. As shown in the introduction, the subsidence in Kunming was highly correlated with urbanization and groundwater extraction, meaning that the degree of urbanization and groundwater extraction with respect to the different times may have led to the different deformation magnitudes [5]. Based on this analysis, the relevant information was investigated and the degree of urbanization and groundwater extraction was found to be extremely intensive in Kunming before 2010 [4]. After this date, Kunming was controlled by the local government, resulting in the slowing down of urbanization and the rise of the groundwater level. In order to further display this investigation, changes in buildings (an index to reflect the degree of urbanization) were extracted through analysis of SAR amplitude and coherence images [37]. Here, changes in buildings mainly refer to new buildings, which were easily identifiable from SAR amplitude and coherence images. The left and right panels of Figure6 show the changes in buildings from January 2007 to March 2011 and June 2011 to January 2016, respectively. The statistics in Figure6 indicate that the area of new buildings was larger in the left panel than the right, demonstrating more intensive urbanization during the ALOS-1 SAR-acquired time. From these investigations, the L-band showed more subsidence than the X-band due to more intensive urbanization and groundwater extraction. The second reason why different subsidence magnitudes were apparent between the L- and X-bands lies in the different imaging geometry, i.e., the L-band, with an incidence angle of 38◦, belongs to the ascending orbit, while the X-band, with incidence angle of 29◦, belongs to the descending orbit. According to the relationship between LOS deformation and three-dimensional (north, east, and up) deformation, the three-dimensional vectors of LOS direction are [– 0.1069 0.6063 0.7880] for the L-band and [0.0842 0.4774 0.8746] for the X-band [38]. As indicated Sensors 2019, 19, x FOR PEER REVIEW 8 of 31 correlated with urbanization and groundwater extraction, meaning that the degree of urbanization and groundwater extraction with respect to the different times may have led to the different deformation magnitudes [5]. Based on this analysis, the relevant information was investigated and the degree of urbanization and groundwater extraction was found to be extremely intensive in Kunming before 2010 [4]. After this date, Kunming was controlled by the local government, resulting in the slowing down of urbanization and the rise of the groundwater level. In order to further display this investigation, changes in buildings (an index to reflect the degree of urbanization) were extracted through analysis of SAR amplitude and coherence images [37]. Here, changes in buildings mainly refer to new buildings, which were easily identifiable from SAR amplitude and coherence images. The left and right panels of Figure 6 show the changes in buildings from January 2007 to March 2011 and June 2011 to January 2016, respectively. The statistics in Figure 6 indicate that the area of new buildings was larger in the left panel than the right, demonstrating more intensive urbanization during the ALOS-1 SAR-acquired time. From these investigations, the L-band showed more subsidence than the X-band due to more intensive urbanization and groundwater extraction. The second reason why different subsidence magnitudes were apparent between the L- and X-bands lies in the different imaging geometry, i.e., the L-band, with an incidence angle of 38, belongs to the ascending orbit, while the X-band, with incidence angle of 29, belongs to the descending orbit. SensorsAccording2019, 19 to, 4425 the relationship between LOS deformation and three-dimensional (north, east, and8 of up) 29 deformation, the three-dimensional vectors of LOS direction are [– 0.1069 0.6063 0.7880] for the L- band and [0.0842 0.4774 0.8746] for the X-band [38]. As indicated by previous studies, the deformation by previous studies, the deformation of Kunming was dominated by the vertical direction and was of Kunming was dominated by the vertical direction and was insignificant in the horizontal direction insignificant in the horizontal direction (north and east) [5–7]. Therefore, the northern and eastern (north and east) [5‒7]. Therefore, the northern and eastern components were directly ignored when components were directly ignored when converting the LOS deformation into the vertical deformation. converting the LOS deformation into the vertical deformation. In this research, the conversion factor In this research, the conversion factor for the L-band was larger than the X-band when connecting the for the L-band was larger than the X-band when connecting the different three-dimensional vectors, different three-dimensional vectors, which may have partly caused the larger subsidence in the L-band which may have partly caused the larger subsidence in the L-band than the X-band. than the X-band.

ALOS-1 COSMO-SkyMed

Xishan Xishan

Guandu Guandu

Changes in buildings Changes in buildings

Water area Water area

FigureFigure 6.6. ChangesChanges inin buildingsbuildings fromfrom ALOS-1ALOS-1 ((left)left) andand COSMO-SkyMedCOSMO-SkyMed ((right)right) in thisthis study.study.

4.2.4.2. ValidationValidation ofof InSARInSAR ResultsResults 4.2.1. Comparison with Leveling-Derived Deformation 4.2.1. Comparison with Leveling-Derived Deformation In order to validate our results, the annual deformation velocities between leveling and InSAR In order to validate our results, the annual deformation velocities between leveling and InSAR were compared, as shown in Figure7, where the left and right figures show the results from the were compared, as shown in Figure 7, where the left and right figures show the results from the L- L-band ALOS-1 and the X-band CSK, respectively. The leveling-measured deformation velocity was band ALOS-1 and the X-band CSK, respectively. The leveling-measured deformation velocity was mainly derived from the local earthquake-monitoring department, so most of these instances were not mainly derived from the local earthquake-monitoring department, so most of these instances were located in severely subsiding areas and therefore show relatively little deformation [4]. To ensure the not located in severely subsiding areas and therefore show relatively little deformation [4]. To ensure comparability of our results, the mean value within a 50 m radius circle with a leveling point center the comparability of our results, the mean value within a 50 m radius circle with a leveling point (as shown by the red stars in Figure2) was extracted from Figure5 using the InSAR observations. The statistics in Figure7 show that the root mean square error (RMSE) values between InSAR and the leveling observations were about 4.5 mm for the L-band and 3.7 mm for the X-band. The higher accuracy for the X-band may have been due to more interferograms being involved in the annual deformation velocity estimation for the X-band (88 interferograms) than L-band (59 interferograms). This comparison is approximately consistent between InSAR and the leveling observations, demonstrating the reliability of our observations in Kunming. Sensors 2019, 19, x FOR PEER REVIEW 9 of 31 center (as shown by the red stars in Figure 2) was extracted from Figure 5 using the InSAR observations. The statistics in Figure 7 show that the root mean square error (RMSE) values between InSAR and the leveling observations were about 4.5 mm for the L-band and 3.7 mm for the X-band. The higher accuracy for the X-band may have been due to more interferograms being involved in the annual deformation velocity estimation for the X-band (88 interferograms) than L-band (59 interferograms). This comparison is approximately consistent between InSAR and the leveling Sensorsobservations,2019, 19, 4425 demonstrating the reliability of our observations in Kunming. 9 of 29

Comparisons between leveling and InSAR observations

(a) (b)

10 10 ALOS results CSK results

0 0

-10 -10

2 -20 R =0.77 -20 R2=0.78

Leveling DeformationLeveling (mm)

Leveling DeformationLeveling (mm) RMSE=4.5 mm RMSE=3.7 mm

-30 -30 -30 -20 -10 0 10 -30 -20 -10 0 10 InSAR Deformation (mm) InSAR Deformation (mm) Figure 7. Comparison of annual deformation velocity between leveling and InSAR from (a) the L-band ALOS-1 Figure 7. Comparison of annual deformation velocity between leveling and InSAR from (a) the L-band and (b) the X-band COSMO-SkyMed. ALOS-1 and (b) the X-band COSMO-SkyMed.

4.2.2.4.2.2. ComparisonComparison ofof DeformationDeformationBetween Between thetheALOS-1 ALOS-1 and and Sentinel-1A Sentinel-1A Datasets Datasets Thirty-oneThirty-one additional additional Sentinel-1A Sentinel-1A imagesimages withwith ascendingascending orbitorbit (Table(Table1 )1) were were processed processed using using the the multi-temporalmulti-temporal InSARInSAR technique to generate generate the the ground ground deforma deformationtion between between 23 23 January January 2015 2015 and and 17 17February February 2017 2017 with with the the purpose purpose of of comparing with ALOS-derived ALOS-derived deformation. deformation. After After setting setting the the spatial-temporalspatial-temporal baselinesbaselines (Figure(Figure A1 in the the Appendix), AppendixA the), the time time series series deformation deformation was was produced produced over overthe study the study area, area,as shown as shown in Figure in Figure A2 in A2 the in Appendix. the Appendix Due toA. no Due overlapping to no overlapping time between time betweenthese two thesedatasets, two datasets, the annual the deformationannual deformation velocity velocity maps maps were were compared compared to investigate to investigate the the relationship, relationship, as asshown shown in in Figure Figure 8a8a,b; and Figure b; Figure8a shows 8a shows the result the result from L-bandfrom L ALOS-1-band ALOS and Figure-1 and 8Figureb shows 8b the shows result the fromresult C-band from C Sentinel-1A.-band Sentinel It was-1A. observed It was observed that there that was there clear was difference clear difference in the distribution in the distribution of subsiding of regionssubsiding between regions these between two results; these twoboth results; Guandu both and Guandu Xishan subsiding and Xishan regions subsiding (black re rectanglesgions (black in Figurerectangles5) in Figurein Figure8a moved 5) in Figure about five8a moved kilometers about to thefive southeast kilometers in Figureto the8 b;southeast the obvious in Figure subsidence 8b; the appearedobvious subsidence in the upper appeared left and bottomin the upper of Figure left8 andb, while bottom this of was Figure not significant 8b, while this in Figure was not8a. Wesignificant think thisin Figure inconsistency 8a. We think is due this to inconsistency the different is SAR-acquired due to the different times, i.e.,SAR the-acquired ALOS-1 times, result i.e., was the derivedALOS-1 betweenresult was January derived 2007 between and March January 2011 2007 while and theMarch Sentinel-1A 2011 while result the wasSentinel derived-1A result between was January derived 2015between and FebruaryJanuary 2015 2017. and This February comparison 2017. This also comparison indicates that also the indicates distribution that the of grounddistribution subsidence of ground in Kunmingsubsidence varied in Kunming dynamically varied with dynamically time. It needs with totime. be emphasizedIt needs to be that emphasized there are notthat suitable there are SAR not imagessuitable with SAR ascending images with orbit ascending during our orbit ALOS-acquired during our ALOS period.-acquired Therefore, period. it isTherefore, difficult to it validateis difficult the to ALOS-derivedvalidate the ALOS ground-derived deformation ground deformation by comparing by withcomparing additional with SARadditional datasets. SAR However, datasets. However, we think thewe ALOSthink the result ALOS was result reliable was when reliable comparing when comparing with the leveling-derivedwith the leveling- deformation.derived deformation.

Sensors 2019, 19, 4425 10 of 29 Sensors 2019, 19, x FOR PEER REVIEW 10 of 31

(a):ALOS-1 Reference point (b): Sentinel-1A Reference point

10 km 10 km

(a) (b)

FigureFigure 8. 8.Comparison Comparison of of deformation deformation betweenbetween the ALOS-1ALOS-1 and and Sentinel Sentinel-1A-1A datasets. datasets. (a ()a T)he The result result of of ALOS-1.ALOS-1 (b. ()b The) The result result of of Sentinel-1A. Sentinel-1A. 4.2.3. Comparison of Deformation Between the COSMO-SkyMed and Sentinel-1A Datasets 4.2.3. Comparison of Deformation Between the COSMO-SkyMed and Sentinel-1A Datasets To further validate the CSK-derived deformation, 60 C-band Sentinel-1A images (Table1) with To further validate the CSK-derived deformation, 60 C-band Sentinel-1A images (Table 1) with descendingdescending orbit orbit were were processed processed usingusing thethe multi-temporal multi-temporal InSAR InSAR technique technique to to generate generate the the time time seriesseries deformation. deformation. This This procedure procedure was was divideddivided into two phases phases due due to to lack lack of of data data between between October October 20162016 and and March March 2018. 2018. TheThe firstfirst phase phase generated generated the the time time series series deformation deformation from from 25 May 25 May2015 2015to 16 to 16September 2016 2016 through through processing processing 18 18 C- C-bandband Sentinel Sentinel-1A-1A images, images, as shown as shown in Figure in Figures A3 and A3 A4and in A4 in the AppendixAppendix. AThe. The second second phase phase generated generated the thetime time series series deformation deformation from from 22 March 22 March 2018 2018 to 1 to 1 SeptemberSeptember 20192019 through processing 42 42 C C-band-band Sentinel Sentinel-1A-1A images, images, as asshown shown in Figure in Figures A5 and A5 A6and in A6 in the Appendix. Appendix AIt .was It was observed observed from from Figure Figures A4 and A4 A6 and that A6 the that Guandu the Guandu and Xishan and subsiding Xishan subsiding regions regionsin Figure in Figure 5 moved5 moved about about five kilometers five kilometers to the to southeast, the southeast, which which was consistent was consistent with Figure with Figure A2. To A2 . Toensure ensure the the consistency consistency of of tim time,e, the the CSK CSK-derived-derived deformation deformation map map spanning spanning from from 15 May 15 May 2015 2015 to 6 to 6 JanuaryJanuary 20162016 (Figure(Figure9 9a)a) waswas comparedcompared withwith Sentinel-derivedSentinel-derived deformation deformation spanning spanning from from 25 25 May May 20152015 to 20to 20 January January 2016 2016 (Figure (Figure9b). 9b). Figure Figure9a 9a was was resampled resampled to to the the geographical geographical coordinate coordinate system system as in Figureas in Figure9b. Unlike 9b. Unlike the last the comparison, last comparison, the distribution the distribution of subsiding of subsiding region was region basically was basically consistent betweenconsistent Figure between9a,b. However,Figure 9a and the slightb. However, di fference the slight was thatdifference the CSK was showed that the a CSK larger showed subsidence a larger than Sentinel-1Asubsidence in than the northSentinel of- Guandu1A in the subsiding north of Guandu region. Thissubsidi phenomenonng region. This may phenomenon have been caused may have by the diffbeenerent caused wavelengths; by the different CSK belongs wavelengths; to the X-band CSK belongs while Sentinel-1Ato the X-band belongs while Sen to thetinel C-band.-1A belongs to the C-band.

Sensors 2019, 19, 4425 11 of 29 Sensors 2019, 19, x FOR PEER REVIEW 11 of 31

(a):COSMO-SkyMed Reference point (b): Sentinel-1A Reference point

10 km 10 km

FigureFigure 9. 9. ComparisonComparison of of deformation deformation between between the theCOSMO COSMO-SkyMed-SkyMed and Sentinel and Sentinel-1A-1A datasets datasets.. The TheFigure Figure 9a9 ashow show the the result result of ofCOSMO COSMO-SkyMed-SkyMed and and Figure Figure 9b show9b show the result the result of Sentinel of Sentinel-1A.-1A.

4.3.4.3. Time Time Series Series Deformation Deformation in in Guandu Guandu AsA showns shown in in Figure Figure5, 5, the the regions regions Guandu Guandu and and Xishan,Xishan, which both both suffer su ffer from from severe severe subsidence, subsidence, werewere observed observed with with the the L- L and- and X-band X-band SAR SAR interferograms. interferograms. To To obtain obtain detailed detailed informationinformation relatingrelating to theseto thesetwo regions, two regions, the time the series time deformation series deformation was estimated was estimated using multi-temporal using multi-temporal InSAR processing, InSAR asprocessing, described in as Sectiondescribed 3.2 in. Figure Section 10 3.2. shows Figure the 10 timeshows series the time deformation series deformation of Guandu of fromGuandu 27 Augustfrom 200727 toAugust 7 March 2007 2011, to 7 March which 2011 was, derivedwhich was from derived L-band from ALOS-1 L-band images, ALOS- and1 images, also showsand also prominent shows subsidenceprominent in subsidence the upper in center the upper of the center figure, of the where figure, the where cumulative the cumulative maximum maximum subsidence subsidence reached reached 110 mm. It was also found that the subsidence presented in the shape of a funnel and 110 mm. It was also found that the subsidence presented in the shape of a funnel and extended to extended to the southeast. The decorrelation in the southeast was due to ongoing construction [4], as the southeast. The decorrelation in the southeast was due to ongoing construction [4], as shown in shown in Figure 6. Figure6. Figure 11 shows the time series deformation of Guandu between 8 August 2011 and 6 January 2016, which was derived from the X-band CSK images. Similarly to Figure 10, prominent subsidence started from the upper center and extended to the southeast. However, most of the subsiding region in the southeast kept high coherence, as shown in Figure 11, while it was decorrelated in Figure 10. This phenomenon can be explained by the fact that most construction during the L-band SAR-acquired time was completed during the X-band SAR-acquired time. The statistics indicate that the cumulative maximum subsidence was 102 mm, which was basically consistent with the L-band result.

Sensors 2019, 19, 4425 12 of 29

Sensors 2019, 19, x FOR PEER REVIEW 12 of 31

Figure 10.FigureTime 10. seriesTime series deformation deformation in in Guandu Guandu (marked(marked with with black black rectangles rectangles in Figure in Figure 5) observed5) observed from from L-band ALOS-1 images. The time series deformation at point P1 will be extracted for further analysis. L-bandSensors ALOS-1 2019, 19, x images. FOR PEER The REVIEW time series deformation at point P1 will be extracted for further13 analysis. of 31 Figure 11 shows the time series deformation of Guandu between 8 August 2011 and 6 January 2016, which was derived from the X-band CSK images. Similarly to Figure 10, prominent subsidence started from the upper center and extended to the southeast. However, most of the subsiding region in the southeast kept high coherence, as shown in Figure 11, while it was decorrelated in Figure 10. This phenomenon can be explained by the fact that most construction during the L-band SAR- acquired time was completed during the X-band SAR-acquired time. The statistics indicate that the cumulative maximum subsidence was 102 mm, which was basically consistent with the L-band result.

Figure 11.FigureTime 11. seriesTime series deformation deformation in in Guandu Guandu (marked(marked with with black black rectangles rectangles in Figure in Figure 5) observed5) observed using X-band COSMO-SkyMed images. The time series deformation at point P1 will be extracted for using X-band COSMO-SkyMed images. The time series deformation at point P1 will be extracted for further analysis. further analysis. To investigate the long-term ground deformation time series, both L-band ALOS-1 and X-band CSK deformation in LOS direction were projected to the vertical direction. After that, the cumulative vertical deformation from the L-band was added to the X-band to connect them. Figure 12 shows the vertical time series deformation of Guandu at point P1 between 27 August 2007 and 6 January 2016. The blue squares and the red circles represent the vertical deformation values for the L-band and the X-band, respectively. It was observed that the deformation at point P1 presented an approximately linear variation with time, particularly for the X-band. The cumulative subsidence reached 209 mm during the period of 2007–2016.

Sensors 2019, 19, 4425 13 of 29

To investigate the long-term ground deformation time series, both L-band ALOS-1 and X-band CSK deformation in LOS direction were projected to the vertical direction. After that, the cumulative vertical deformation from the L-band was added to the X-band to connect them. Figure 12 shows the vertical time series deformation of Guandu at point P1 between 27 August 2007 and 6 January 2016. The blue squares and the red circles represent the vertical deformation values for the L-band and the X-band, respectively. It was observed that the deformation at point P1 presented an approximately linear variation with time, particularly for the X-band. The cumulative subsidence reached 209 mm duringSensors 2019 the, 19 period, x FOR ofPEER 2007–2016. REVIEW 14 of 31

Figure 12. Time series deformation at point P1 marked in Figures 10 and 11. The blue squares and red Figure 12. Time series deformation at point P1 marked in Figure 10 and Figure 11. The blue squares circles represent deformation for the L-band and the X-band, respectively. and red circles represent deformation for the L-band and the X-band, respectively. 4.4. Time Series Deformation in Xishan 4.4. Time Series Deformation in Xishan Like Guandu, Xishan is another region with severe subsidence. Figure 13 displays the time series deformationLike Guandu, of Xishan Xishan from is another 27 August region 2007 with to 7severe March subsidence. 2011 derived Figure from 13 L-band displays ALOS-1 the time images. series Compareddeformation with of X Guandu,ishan from subsidence 27 August was 2007 not to significant 7 March in2011 Xishan, derived with from the L cumulative-band ALOS deformation-1 images. rangingCompared from with 60Guandu, mm to subsidence 10 mm during was the not ALOS-1 significant SAR-acquired in Xishan, with time. the The cumul subsidenceative deformation presented − inranging a long from rectangle − 60 mm shape to 10 and mm extendedduring the southwest ALOS-1 SAR in Xishan,-acquired which time. wasThe subsidence different to presented Guandu. Furtherin a longobservation rectangle shape indicated and extended that the southwest subsiding in points Xishan, in which Xishan was were different not as to concentrated Guandu. Further as in Guandu,observation showing indicated a certain that discreteness.the subsiding This points phenomenon in Xishan waswere related not as to concentrated the distribution as in of Guandu, high-rise buildingsshowing a[29 certain]. discreteness. This phenomenon was related to the distribution of high-rise buildings [29].

Sensors 2019, 19, 4425 14 of 29 Sensors 2019, 19, x FOR PEER REVIEW 15 of 31

Figure 13.13. TimeTime series series deformation deformation in in Xishan Xishan (marked (marked with with black black rectangles rectangles in Figure in Figure5) observed 5) observed using L-bandusing L ALOS-1-band ALOS images.-1 images. The time The series time deformation series deformation at point at P2 point willbe P2 extracted will be extracted for further for analysis. further analysis. The time series deformation of Xishan from 8 August 2011 to 6 January 2016 was observed from the X-bandThe time CSK series images, deformation as shown of in Xishan Figure from 14. It 8 wasAugust found 2011 that to the6 January point targets2016 was in Figureobserved 14 werefrom denserthe X-band than CSK those images, in Figure as shown 13, which in Figure may have14. It benefitedwas found from that the the higherpoint targets spatial in resolution Figure 14 of were the X-banddenser than than those the L-band in Figure [18 ].13, Similar which tomay Figure have 13 benefited, the subsidence from the was higher approximately spatial resolution in the shapeof the X of- aband long than rectangle the L- thatband extended [18]. Similar to the to southwest.Figure 13, the However, subsidence the obviouswas approximately subsidence in appeared the shape in of the a upperlong rectangle left in Figure that 14extended, while this to the was southwest. not significant However, in Figure the 13 obvious. This di subsidencefference was appeared due to the in new the completedupper left in constructions, Figure 14, while which this was was confirmed not significant by filed in Figure investigation 13. This [4 difference]. The statistics was due show to the that new the cumulativecompleted constructions, maximum subsidence which was was confirmed 52 mm duringby filed this investigation period, which [4]. The was statistics slightly show less than that thethe L-bandcumulative result. maximum subsidence was 52 mm during this period, which was slightly less than the L- band result.

Sensors 2019, 19, 4425 15 of 29 Sensors 2019, 19, x FOR PEER REVIEW 16 of 31

Figure 14. Time series deformation in Xishan (marked with black rectangles in Figure 55)) observedobserved from X-bandX-band COSMO-SkyMedCOSMO-SkyMed images. The time series deformation at point P2 will be extracted for further analysis.

Figure 1515 showsshows thethe verticalvertical time time series series deformation deformation of of Xishan Xishan at at point point P2 P2 from from 27 27 August August 2007 2007 to 6to January6 January 2016, 2016 where, where the the blue blue squares squares and and red red circles circles represent represent thethe verticalvertical deformationdeformation for the L-bandL-band and the X-band,X-band, respectively. This This was was similar similar to to that observed in Figure 12,12, in that the cumulative deformation from the L L-band-band was added to the X-bandX-band result after converting the LOS deformation into into the the vertical vertical deformation. deformation. Figure Figure 15 shows 15 shows that the that deformation the deformation at point P2 at presented point P2 anpresented approximately an approximately linear variation linear with variation time, wherewith time, the cumulativewhere the cumulative subsidence subsidence was 98 mm was during 98 mm the periodduring ofthe 2007–2016. period of 2007–2016.

Sensors 2019, 19, x FOR PEER REVIEW 17 of 31 Sensors 2019, 19, 4425 16 of 29 Sensors 2019, 19, x FOR PEER REVIEW 17 of 31

Figure 15. Time series deformation at point P2 marked in Figure 13 and Figure 14. Figure 15. Time series deformation at point P2 marked in Figures 13 and 14. Figure 15. Time series deformation at point P2 marked in Figure 13 and Figure 14. 5.5. DiscussionDiscussion 5. Discussion 5.1.5.1. SubsidenceSubsidence DueDue toto SoftSoft SoilSoil ConsolidationConsolidation 5.1. Subsidence Due to Soft Soil Consolidation KunmingKunming waswas aa typicaltypical lacustrinelacustrine sediment basin that was filled filled by a large area of soft soil [[1].1]. FigureFigureKunming 16 16a,a, which which was displays a displaystypical the lacustrine changes the changes in sediment the shoreline in the basin shoreline of that Lake was Dian, of filled Lake clearly by Dian, a indicates large clearly area the of indicatesdevelopment soft soil [1]. the ofFiguredevelopment Kunming 16a, basin. whichof Kunming In displays this context,basin. the In changes thethis stratum context, in the of the Kunming shoreline stratum wasof of Kunming mainlyLake Dian, composed was clearlymainly of indicatescomposed clay, muddy the of clay,developmentclay, andmuddy peaty clay,of soils. Kunming and These peaty basin. soft soils. soils In These hadthis manycontext,soft soils unfavorable the had stratum many properties unfavorableof Kunming regarding propertieswas mainly their regarding use composed in projects, their of suchclay,use in asmuddy projects, high compressibility,clay, such and as peatyhigh compressibility, rheology,soils. These and soft thixotropy rheology, soils had [ and4 many]. Subsidence thixotropy unfavorable easily[4]. Subsidence properties occurred whenregardingeasily soilsoccurred their had highusewhen in compressibility projects, soils had such high as due compressibility high to compressibility, the natural due consolidation to rheology, the natural ofand the consolidationthixotropy soils and [4]. man-made of Subsidence the soils building andeasily man loadoccurred-made [39]. Figurewhenbuilding soils16 bload shows had [39]. high the Figure stratigraphic compressibility 16b shows profile the due stratigraphic along to the the natural line pr ofofile consolidation CD along (black the solid line of line theof CD in soils Figure(black and 16 solid mana), which-linemade in correspondsbuildingFigure 16a), load towhich [39]. the Figure position corresponds 16b of shows boreholes. to thethe positionstratigraphic It was of found boreholes profile that along both. It was Xishanthe line found andof CD that Guandu ( black both solid Xishan were line filled and in withFigureGuandu di ff 16a),erent were which layersfilled with correspondsof soft different soils with to layers the diff positionerentof soft quaternary soils of boreholeswith thicknesses.different. It was quaternary Based foundon thatthicknesses. such both a stratigraphic Xishan Based and on structure,Guandusuch a str were itatigraphic was filled deduced with structure, that different soft it soil waslayers consolidation deduced of soft soils that was withsoft one different soilof the consolidation factors quaternary causing was thicknesses. the one severe of subsidence the Based factors on observedsuchcausing a strthe inatigraphic thesevere study subsidence structure, area. observed it was deduced in the study that area.soft soil consolidation was one of the factors causing the severe subsidence observed in the study area. (a) (b)

(a) (b)

C D C D

Coverage of Lake Dian around 2016

CoverageCoverage of of Lake Lake Dian Dian around around 2016 766 Coverage of Lake Dian around 2000766 BC Coverage of Lake Dian around 2000 BC FigureFigure 16.16. ((aa)) ChangesChanges in thethe shorelineshoreline ofof LakeLake Dian;Dian; (b)) stratigraphicstratigraphic profileprofile alongalong thethe lineline ofof CD,CD, whichFigurewhich corresponds corresponds16. (a) Changes totothe the inposition positionthe shorelineof ofboreholes. boreholes of Lake. Dian; (b) stratigraphic profile along the line of CD, which corresponds to the position of boreholes.

Sensors 2019, 19, 4425 17 of 29 Sensors 2019, 19, x FOR PEER REVIEW 18 of 31

The relevant materials show that the magnitude of subsidence may be related to the quaternary The relevant materials show that the magnitude of subsidence may be related to the quaternary sediment thickness [40]. Thus, the relationship between ground deformation and quaternary sediment thickness [40]. Thus, the relationship between ground deformation and quaternary sediment sediment thickness was investigated and analyzed over the study area. The quaternary sediment thickness was investigated and analyzed over the study area. The quaternary sediment thickness thickness of Kunming was collected from the local geological department, then the spatial of Kunming was collected from the local geological department, then the spatial distributions of distributions of the deformation and the quaternary sediment thickness were compared, as shown in the deformation and the quaternary sediment thickness were compared, as shown in Figure 17a. Figure 17a. The comparison results show the approximate consistency between the ground The comparison results show the approximate consistency between the ground deformation and deformation and the quaternary sediment thickness in the spatial distribution; the greater the the quaternary sediment thickness in the spatial distribution; the greater the subsidence, the thicker subsidence, the thicker the quaternary, particularly for the subsiding region of Xishan, where the the quaternary, particularly for the subsiding region of Xishan, where the sediment thickness was sediment thickness was 500–800 m [4]. To further investigate this correlation, regression analysis was 500–800 m [4]. To further investigate this correlation, regression analysis was carried out for ground carried out for ground deformation and quaternary sediment thickness at the selected sampling deformation and quaternary sediment thickness at the selected sampling points. In order to ensure points. In order to ensure the rationality of comparison, points with different quaternary sediment the rationality of comparison, points with different quaternary sediment thickness were randomly thickness were randomly selected in this study, as shown in Figure 17a. After collecting the selected in this study, as shown in Figure 17a. After collecting the information of quaternary sediment information of quaternary sediment thickness, the mean value within a 50 m radius circle with a thickness, the mean value within a 50 m radius circle with a sampling point center was extracted sampling point center was extracted from L-band ALOS-derived deformation as the InSAR from L-band ALOS-derived deformation as the InSAR observations, as shown in Figure 17b [40]. observations, as shown in Figure 17b [40]. It was found that some large subsidence points It was found that some large subsidence points corresponded to the thick quaternary sediment, while corresponded to the thick quaternary sediment, while some small subsidence points corresponded some small subsidence points corresponded to the shallow quaternary sediment. The statistics show to the shallow quaternary sediment. The statistics show that the correlation coefficient between these that the correlation coefficient between these two factors is about 0.65, suggesting that there is two factors is about −0.65, suggesting that there is somewhat negative− correlation between ground somewhat negative correlation between ground deformation and quaternary sediment thickness. deformation and quaternary sediment thickness. When ignoring the other factors (e.g., building load, When ignoring the other factors (e.g., building load, construction), our result indicates some degree construction), our result indicates some degree of correlation between the ground deformation and of correlation between the ground deformation and soft soil consolidation. However, it was also soft soil consolidation. However, it was also observed that there were some inconsistent points, e.g., observed that there were some inconsistent points, e.g., the large subsidence but shallow sediment the large subsidence but shallow sediment seen in Figure 17b. This inconsistency may be due to the seen in Figure 17b. This inconsistency may be due to the heterogeneous thickness of the compressible heterogeneous thickness of the compressible soils across the quaternary deposits or due to the other soils across the quaternary deposits or due to the other factors causing the severe subsidence observed factors causing the severe subsidence observed in the study area [40]. Based on the analyses of the in the study area [40]. Based on the analyses of the stratigraphic structure in Kunming (Figure 16) stratigraphic structure in Kunming (Figure 16) and the relationship between ground deformation and and the relationship between ground deformation and quaternary sediment thickness (Figure 17), quaternary sediment thickness (Figure 17), it was deduced that unconsolidated sedimentary deposits it was deduced that unconsolidated sedimentary deposits was one of the factors causing the severe was one of the factors causing the severe subsidence in Kunming. subsidence in Kunming.

(a) (b) Relationship between subsidence and sediment thickness

R2=-0.65

Sampling points of quaternary thickness

FigureFigure 17. 17. (a) Quaternary sediment thicknessthickness contours contours in in Kunming Kunming and and ( (bb)) relationship relationship between between groundground deformationdeformation andand quaternaryquaternary sedimentsediment thickness.thickness.

ForFor furtherfurther investigation of the relationship between ground deformation and geological setting, thethe quaternary active active faults faults were were superimposed superimposed on on the the L- L-bandband ALOS ALOS-derived-derived deformation deformation map, map, as asshown shown in Figure in Figure 18, where18, where the black the black solid solidline represents line represents the qua theternary quaternary active faults. active It faults. was observed It was observedthat there that were there eight were faults eight around faults around Kunming, Kunming, which which controlled controlled the formation the formation and and development development of the basin. As many studies have indicated, the spatial distribution and shape of subsidence might be related with the faults [19−21]. In this study, Guandu subsiding region presented in the shape of a

Sensors 2019, 19, 4425 18 of 29

ofSensors the basin.2019, 19,As x FOR many PEER studies REVIEW have indicated, the spatial distribution and shape of subsidence might19 of 31 be related with the faults [19–21]. In this study, Guandu subsiding region presented in the shape of a narrow funnel and extendedextended inin anan approximateapproximate south-northsouth-north direction,direction, whichwhich followedfollowed thethe generalgeneral trendstrends of of the the surrounding surrounding F149 Fand149 and F150 Ffaults.150 faults. The TheXishan Xishan subsiding subsiding region region presented presen in a longted in rectangle a long shaperectangle and shape extended and to extended the southwest, to the which southwest, didn’t follow which the didn’t general follow trends the of general the surrounding trends of F the151, Fsurrounding152, and F55 Ffaults.151, F152 However,, and F55 faults. careful However, inspection careful suggests inspection that Xishan suggests subsiding that Xishan region subsiding might be boundedregion might by these be bounded three faults: by these theeast three was faults: bounded the east by was F152 ,bounded the west by and F152 south, the were west boundedand south by were F55, andbounded F152 acted by F55, as and barrier F152 acted to impede as barrier the horizontal to impede propagation the horizontal of deformationpropagation onof deformation both sides. Based on both on thissides. analysis, Based on there this was analysis, a certain there degree was a of certain correlation degree between of correlation ground between deformation ground and deformation quaternary activeand quaternary faults over active the area. faults over the area.

Figure 18.18. The relationship between ground deformation and quaternary active faults in Kunming. TheThe blackblack solidsolid lineline showsshows thethe quaternaryquaternary activeactive faults.faults.

5.2. Subsidence Due to High-Rise Building Load 5.2. Subsidence Due to High-Rise Building Load As introducedintroducedin in Section Section2 ,2, Kunming Kunming is is surround surround by by mountains mountains to theto the north, north, west, west, and and east, east, and and by aby lake a lake to theto the south. south. Under Under these these condition, condition, available available space space is veryis very limited limited and and rapid rapid urbanization urbanization in Kunmingin Kunming is inevitable. is inevitable. To solve To solve this problem,this problem, a large a numberlarge number of high-rise of high buildings-rise buildings with high with building high densitybuilding were density set up were around set Lakeup around Dian over Lake the Dian last fewover decades. the last Figure few decades. 19 shows Figure the changes 19 shows in land the coverchanges from in land 1984 cover to 2016, from where 1984 the to yellow2016, where curve the indicates yellow thecurve main indicates range ofthe changes. main range It is of clear changes. that a largeIt is clear area that of green a large land area was of occupied green land by was high-rise occupied buildings by high and-rise other buildings structures. and Thisother change structures. was moreThis change obvious was in Xishanmore obvious and Guandu, in Xishan which and corresponded Guandu, which to the corresponded severely subsiding to the severely regions, subsiding as shown inregions, Figure as20 . shown Considering in Figure the soft 20. Consideringsoil foundations the in soft Figure soil 16 foundations, this high-rise in Figure building 16, load this inevitably high-rise causedbuilding the load severe inevitably subsidence caused in Kunming, the severe particularly subsidence for in Xishan Kunming, and Guandu particularly [41,42 for]. Xishan and Guandu [41,42].

Sensors 2019, 19, 4425 19 of 29 Sensors 2019, 19, x FOR PEER REVIEW 20 of 31

Figure 19. Changes in land cover over Kunming from (a) 1984 to ( b) 2016; the yellow curve shows the main changes.

To further investigate the relationship between subsidence and building load, the building densities over 35 sampling points were extracted and analyzed over Kunming. Building density is the floor area of the building divided by the total area of the site [43]. It should be noted here that the floor area of a building only refers to the single floor area and doesn’t contain the height information. The information of 35 sampling points was collected through filed investigation organized in June, 2016. In order to ensure the rationality of comparison, buildings with different densities were investigated and analyzed in this study. The selected sampling points were mainly concentrated in the regions sensitive to ground deformation, such as residential areas with high-rise buildings, government administrative areas, dense business districts, as well as important infrastructure regions, as shown in Figure 21a. After collecting the information of building densities, the mean value within the investigated area was extracted from L-band ALOS-derived deformation as the InSAR observations, as shown in Figure 21b. The statistics show that the correlation coefficient between ground deformation and building density was about 0.61, suggesting somewhat negative correlation between them. When ignoring the other indexes (e.g., − building height) and assuming that building density was the only index to reflect the building load, our result indicates some degree of correlation between the ground deformation and building load. However, it was also observed that there were some inconsistent points, e.g., the large subsidence but small density seen in Figure 20. This inconsistency may be due to the other factors causing the severe subsidence. Based on the analyses of land cover change in Kunming (Figures 19 and 20) and the relationship between ground deformation and building density (Figure 21), it was deduced that building load was one of the factors causing the severe subsidence in Kunming, particularly for Xishan and Guandu.

Sensors 2019, 19, 4425 20 of 29 Sensors 2019, 19, x FOR PEER REVIEW 21 of 31

(a) (b)

Xishan-1984 Xishan-2016

(c) (d)

Guandu-1984 Guandu-1984

Figure 20. The changes in land cover over Xishan from ( a) 1984 to ( b) 2016, and and over Guandu Guandu from ( c) 19841984 toto ((dd)) 2016.2016.

To further investigate the relationship between subsidence and building load, the building densities over 35 sampling points were extracted and analyzed over Kunming. Building density is the floor area of the building divided by the total area of the site [43]. It should be noted here that the floor area of a building only refers to the single floor area and doesn’t contain the height information. The information of 35 sampling points was collected through filed investigation organized in June, 2016. In order to ensure the rationality of comparison, buildings with different densities were investigated and analyzed in this study. The selected sampling points were mainly concentrated in the regions sensitive to ground deformation, such as residential areas with high-rise buildings, government administrative areas, dense business districts, as well as important infrastructure regions, as shown in Figure 21a. After collecting the information of building densities, the mean value within the investigated area was extracted from L-band ALOS-derived deformation as the InSAR observations, as shown in Figure 21b. The statistics show that the correlation coefficient between ground deformation and building density was about −0.61, suggesting somewhat negative correlation between them. When ignoring the other indexes (e.g., building height) and assuming that building density was the only index to reflect the building load, our result indicates some degree of correlation between the ground deformation and building load. However, it was also observed that

Sensors 2019, 19, x FOR PEER REVIEW 22 of 31 there were some inconsistent points, e.g., the large subsidence but small density seen in Figure 20. This inconsistency may be due to the other factors causing the severe subsidence. Based on the analyses of land cover change in Kunming (Figures 19 and 20) and the relationship between ground deformation and building density (Figure 21), it was deduced that building load was one of the Sensorsfactors2019 causing, 19, 4425 the severe subsidence in Kunming, particularly for Xishan and Guandu. 21 of 29

(a) (b)

Sampling points of building densities

Figure 21. The distribution of sampling points of building densities (a) and the relationshiprelationship betweenbetween ground deformation andand buildingbuilding densitydensity ((bb).). 5.3. Subsidence Due to Groundwater Exploitation 5.3. Subsidence Due to Groundwater Exploitation Groundwater is the main water supply in Kunming, where the average daily groundwater Groundwater is the main water supply in Kunming, where the average daily groundwater exploitation was more than 170,000 m3 prior to 2010 [4]. Figure 22a shows the distribution of exploitation was more than 170,000 m3 prior to 2010 [4]. Figure 22a shows the distribution of groundwater exploited regions in Kunming, which corresponds to the groundwater supplying sources. groundwater exploited regions in Kunming, which corresponds to the groundwater supplying It was found that exploited regions mainly lie in eastern Guandu and Chenggong, northern Longtoujie, sources. It was found that exploited regions mainly lie in eastern Guandu and Chenggong, northern and western Majie. Among them, Guandu, Chenggong, and Majie belong to the karst groundwater, Longtoujie, and western Majie. Among them, Guandu, Chenggong, and Majie belong to the karst and Longtoujie belongs to the pore groundwater. According to the observations of groundwater groundwater, and Longtoujie belongs to the pore groundwater. According to the observations of levels, eastern Guandu declined 12.1 m between 2004 and 2012, which was the greatest decline in groundwater levels, eastern Guandu declined 12.1 m between 2004 and 2012, which was the greatest Kunming, as shown in the blue line of Figure 22b [44]. It was found that there was a dramatic decline decline in Kunming, as shown in the blue line of Figure 22b [44]. It was found that there was a between 2006 and 2010 and, subsequently, a slight decline between 2010 and 2012. This variation dramatic decline between 2006 and 2010 and, subsequently, a slight decline between 2010 and 2012. was consistent with the time that restrictions were placed on groundwater extraction. As many This variation was consistent with the time that restrictions were placed on groundwater extraction. pioneer studies indicated, groundwater exploitation certainly resulted in ground subsidence [45,46]. As many pioneer studies indicated, groundwater exploitation certainly resulted in ground Therefore, severe subsidence was observed in Guandu, as shown in the red line of Figure 22b, where subsidence [45,46]. Therefore, severe subsidence was observed in Guandu, as shown in the red line the cumulative subsidence reached 100 mm during the period of 2007–2011. Comparison between of Figure 22b, where the cumulative subsidence reached 100 mm during the period of 2007–2011. the change of groundwater level and ground deformation in Figure 22 indicates that they presented Comparison between the change of groundwater level and ground deformation in Figure 22 indicates similar variations in temporal domain. For the other exploited regions, the changes of groundwater that they presented similar variations in temporal domain. For the other exploited regions, the level weren’t as significant as Guandu, and ground deformation was also not as severe, as shown changes of groundwater level weren’t as significant as Guandu, and ground deformation was also in Figure 22c–e. Based on the analyses of groundwater exploited regions and level changes, it was not as severe, as shown in Figure 22c‒e. Based on the analyses of groundwater exploited regions and deduced that groundwater exploitation was one of factors causing the severe subsidence in Kunming, level changes, it was deduced that groundwater exploitation was one of factors causing the severe particularly for Guandu. subsidence in Kunming, particularly for Guandu.

Sensors 2019, 19, x FOR PEER REVIEW 23 of 31 Sensors 2019, 19, 4425 22 of 29

(a)

Longtou

Guandu

Majie Chenggong

Groundwater over-exploited region with karst water

Groundwater over-exploited region with pore water

10 km

(b) (c)

Guandu Chenggong

(d) (e)

Longtoujie Majie

Figure 22. The distribution of groundwater exploited regions in Kunming (a) and relationship between annualFigure changes 22. The of groundwater distribution level of groundwater and ground exploited deformation regions in Guandu in Kunming (b), Chenggong (a) and relationship (c), Longtoujie (d), andbetween Majie annual (e). The changes ground of groundwater deformation level is derived and ground from deformation L-band ALOS-1. in Guandu (b), Chenggong (c), Longtoujie (d), and Majie (e). The ground deformation is derived from L-band ALOS-1. 6. Conclusions 6.With Conclusion the development of urbanization, ground subsidence has become one of the most prominent geologicalWith hazards the development in Kunming andof urbanization, has caused substantial ground subsidence damage tohas homes, become roads, one canals, of the pipelines,most andprominent other types geological of infrastructure. hazards in In Kunming this study, and a decade has caused of time substantial series deformation damage to homes, in Kunming roads, was derivedcanals, using pipelines, a multi-temporal and other types InSAR of infrastructure. technique to In provide this study, scientific a decade evidence of time series regarding deformation the sound managementin Kunming of urbanization. was derived using Based a multi on this-temporal study, the InSAR following technique conclusions to provide were scientific made. evidence (1) Two severe subsiding regions were observed using L- and X-band SAR data. Using 20 L-band

ALOS-1 images and 40 X-band COSMO-SkyMed images, the annual deformation velocity and time series deformation maps of Kunming were retrieved through multi-temporal InSAR processing. The results showed that most areas of Kunming are in a state of subsidence and severely subsiding Sensors 2019, 19, 4425 23 of 29 regions are concentrated in Guandu and Xishan, where the cumulative subsidence reached 210 mm and 98 mm during the period of 2007–2016, respectively. Comparisons of ground deformation between InSAR and leveling observations indicate that RMSE values are about 4.5 mm for the L-band and 3.7 mm for the X-band, demonstrating the reliability of our observations. (2) A correlation analysis was conducted to investigate the causes of the observed subsidence. The results show that the severe subsidence in Kunming might be the result of soft soil consolidation, building load, and groundwater extraction. Among these factors, soft soil was the foundation for the subsidence, and the building load and groundwater extraction factors accelerated the subsidence process. This study contributes to the understanding of the spatial-temporal evolution of ground deformation in Kunming. However, it is difficult to clearly describe the mechanism of land subsidence due to a lack of related materials, such as continuous ground-based deformation monitoring data. Therefore, sufficient in situ measurements will be collected in the future to analyze the mechanism of ground deformation in Kunming.

Author Contributions: Conceptualization, Q.Z.; funding acquisition, Q.Z.; investigation, W.Z. and W.-L.L.; validation, Y.Y.; visualization, W.-L.L.; writing—original draft, W.Z.; writing—review and editing, Y.Z., W.Q., and C.-S.W. All authors conceived the study and reviewed and approved the manuscript. Funding: This research was funded by Chang’an University (Xi’an, China) through the National Key R&D Program of China (2018YFC1505101), Natural Science Foundation of China projects (NSFC) (No: 41790445, 41731066, 41674001, 41604001), Natural Science Basic Research Plan in Shaanxi Province of China (2018JQ4031, 2019JM-202), and the Fundamental Research Funds for the Central Universities, CHD (No: 300102269207, 300102269303, 300102268204). This research was also funded by the Open Fund of Key Laboratory of Urban Land Resources Monitoring and Simulation, Ministry of Land and Resource (No. KF-2018-03-004). Acknowledgments: Several figures were prepared using Generic Mapping Tools software. We are grateful to the anonymous reviewers for their constructive comments to improve this manuscript. ConflictsSensors 2019 of Interest:, 19, x FORThe PEER authors REVIEW declare no conflict of interest. 25 of 31

AppendixAppendix A

FigureFigure A1. A1. TheThe distribution distribution of the of temporal the temporal and spatial and spatialbaselines baselines from Sentinel from-1A Sentinel-1A with ascending with ascendingorbit between 23 orbitJanuary between 2015 and 23 January17 February 2015 2017. and 17 February 2017.

Sensors 2019, 19, 4425 24 of 29 Sensors 2019, 19, x FOR PEER REVIEW 26 of 31

FigureFigure A2. A2. TimeTime series series deformation deformation in in Kunming Kunming observed observed using using C- C-bandband Sentinel Sentinel-1A-1A images images with with ascendingascending orbit orbit between between 23 January 23 January 2015 2015and 17 and February 17 February 2017. 2017.

Sensors 2019, 19, 4425 25 of 29 Sensors 2019, 19, x FOR PEER REVIEW 27 of 31

FigureFigure A3. A3. TheThe distribution distribution of of the the temporal temporal and and spatial spatial baselines baselines from from Sentinel-1A Sentinel-1A with with descending descending orbit orbitbetween between 25 May 25 2015 May and 2015 16 and September 16 September 2016. 2016.

FigureFigure A4. A4. TimeTime series series deformation deformation in in Kunming Kunming observed observed using using C-band C-band Sentinel-1A Sentinel-1A images images with with descendingdescending orbit between between 25 25 May May 2015 2015 and and 16 16 September September 2016. 2016.

Sensors 2019, 19, 4425 26 of 29 Sensors 2019, 19, x FOR PEER REVIEW 28 of 31

Figure A5. The distribution of the temporal and spatial baselines from Sentinel-1A with descending orbit Figure A5. Thebetween distribution 22 March 2018 of and the 1 September temporal 2019.and spatial baselines from Sentinel-1A with descending orbit between 22Sensors March 2019, 19, 2018x FOR PEER and REVIEW 1 September 2019. 29 of 31

Figure A6. Time series deformation in Kunming observed using C-band Sentinel-1A images with Figure A6. Time seriesdescending deformation orbit between 22 March in Kunming 2018 and 1 September observed 2019. using C-band Sentinel-1A images with descending orbit between 22 March 2018 and 1 September 2019.

Sensors 2019, 19, 4425 27 of 29

References

1. Fu, B.C.; Huang, Y.; Li, Q.S.; Fu, J.Q. Research of Cause and Classification of the Shallow Layer Soft Soil for Kunming Basin. J. Kunming Univ. Sci. Technol. 2000, 25, 22–26. 2. Lu, N.; Hernandez, A.J.; Ramsey, R.D. Land cover dynamics monitoring with Landsat data in Kunming, China: A cost-effective sampling and modelling scheme using Google Earth imagery and random forests. Geocarto Int. 2015, 30, 186–201. [CrossRef] 3. Hu, R.L.; Yue, Z.Q.; Wang, L.U.; Wang, S.J. Review on current status and challenging issues of land subsidence in China. Eng. Geol. 2004, 76, 65–77. [CrossRef] 4. Yang, Y.; Huang, X.M.; Zhang, Q.; Zhu, W. Investigation Report of Surface Cover and Ground Deformation of Kunming, China (In Chinese); Yunnan Institute of Surveying and Mapping Engineering: Kunming, China, 2017. 5. Youxin, J.C.; Ruiqi, H.Y. Development Process and Characteristics of Land Subsidence in Kunming. J. Seismol. Res. 2001, 24, 55–60. 6. Chuandong, X.; Shucheng, T.; Feng, L. Clay minerals in Quaternary clayey soil and its relation to the land subsidence in Kunming basin area. Acta Petrologica et Mineralogica 2001, 20, 437–440. 7. Chuandong, X.; Xing, L.; Baozhu, L. Mechanism analysis of land subsidence in Kunming city area. Chin. J. Geol. Hazard Control 2004, 15, 47654. 8. Yu, C.; Li, Z.; Chen, J.; Hu, J.C. Small magnitude co-seismic deformation of the 2017 Mw 6.4 Nyingchi earthquake revealed by InSAR measurements with atmospheric correction. Remote Sens. 2018, 10, 684. [CrossRef] 9. Jo, M.J.; Jung, H.S.; Won, J.S. Detecting the source location of recent summit inflation via three-dimensional InSAR observation of K¯ılauea volcano. Remote Sens. 2015, 7, 14386–14402. [CrossRef] 10. Anantrasirichai, N.; Biggs, J.; Albino, F.; Bull, D. A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets. Remote Sens. Environ. 2019, 230, 111179. [CrossRef] 11. Valade, S.; Ley, A.; Massimetti, F.; D’Hondt, O.; Laiolo, M.; Coppola, D.; Loibl, D.; Hellwich, O.; Walter, T.R. Towards Global Volcano Monitoring Using Multisensor Sentinel Missions and Artificial Intelligence: The MOUNTS Monitoring System. Remote Sens. 2019, 11, 1528. [CrossRef] 12. Yu, H.; Lee, H.; Cao, N.; Lan, Y. Optimal baseline design for multibaseline InSAR phase unwrapping. IEEE Trans. Geosci. Remote Sens. 2019, 57, 5738–5750. [CrossRef] 13. Prébet, R.; Yan, Y.; Jauvin, M.; Trouvé, E. A Data-Adaptive EOF-Based Method for Displacement Signal Retrieval from InSAR Displacement Measurement Time Series for Decorrelating Targets. IEEE Trans. Geosci. Remote Sens. 2019, 57, 5829–5852. [CrossRef] 14. Wnuk, K.; Walton, G.; Zhou, W. Four-dimensional filtering of InSAR persistent scatterers elucidates subsidence induced by tunnel excavation in the Sri Lankan highlands. J. Appl. Remote Sens. 2019, 3, 34508. [CrossRef] 15. Zou, W.; Chen, L. Determination of Optimum Tie Point Interval for SAR Image Coregistration by Decomposing Autocorrelation Coefficient. IEEE Trans. Geosci. Remote Sens. 2019, 5067–5084. [CrossRef] 16. Li, Z.; Cao, Y.; Wei, J.; Duan, M.; Wu, L.; Hou, J.; Zhu, J. Time-series InSAR ground deformation monitoring: Atmospheric delay modeling and estimating. Earth-Sci. Rev. 2019, 192, 258–284. [CrossRef] 17. Wu, S.; Zhang, L.; Ding, X.; Perissin, D. Pixel-Wise MTInSAR Estimator for Integration of Coherent Point Selection and Unwrapped Phase Vector Recovery. IEEE Trans. Geosci. Remote Sens. 2018, 57, 59–68. [CrossRef] 18. Zhang, Q.; Zhu, W.; Ding, X.; Zhao, C.; Yang, C.; Qu, W. Two-dimensional deformation monitoring over Qingxu (China) by integrating C-, L-and X-bands SAR images. Remote Sens. Lett. 2014, 5, 27–36. [CrossRef] 19. Haghighi, M.H.; Motagh, M. Ground surface response to continuous compaction of aquifer system in Tehran, Iran: Results from a long-term multi-sensor InSAR analysis. Remote Sens. Environ. 2019, 221, 534–550. [CrossRef] 20. Peng, M.; Zhao, C.; Zhang, Q.; Lu, Z.; Li, Z. Research on Spatiotemporal Land Deformation (2012–2018) over Xi’an, China, with Multi-Sensor SAR Datasets. Remote Sens. 2019, 11, 664. [CrossRef] 21. Sarychikhina, O.; Glowacka, E.; Robles, B. Multi-sensor DInSAR applied to the spatiotemporal evolution analysis of ground surface deformation in Cerro Prieto basin, Baja California, Mexico, for the 1993–2014 period. Nat. Hazards 2018, 92, 225–255. [CrossRef] Sensors 2019, 19, 4425 28 of 29

22. Aimaiti, Y.; Yamazaki, F.; Liu, W. Multi-sensor InSAR analysis of progressive land subsidence over the Coastal City of Urayasu, Japan. Remote Sens. 2018, 10, 1304. [CrossRef] 23. Solari, L.; Del, M.; Bianchini, S.; Ciampalini, A.; Ezquerro, P.; Montalti, R.; Raspini, F.; Moretti, S. From ERS 1/2 to Sentinel-1: Subsidence monitoring in Italy in the last two decades. Front. Earth. Sci. 2018, 6, 149. [CrossRef] 24. Aslan, G.; Cakır, Z.; Ergintav, S.; Lasserre, C.; Renard, F. Analysis of secular ground motions in istanbul from a long-term insar time-series (1992–2017). Remote Sens. 2018, 10, 408. [CrossRef] 25. Rimba, A.B.; Osawa, T.; Parwata, I.N.S.; As-syakur, A.R.; Kasim, F.; Astarini, I.A. Physical assessment of coastal vulnerability under enhanced land subsidence in Semarang, Indonesia, using multi-sensor satellite data. Adv. Space Res. 2018, 61, 2159–2179. 26. Fuhrmann, T.; Garthwaite, M.C. Resolving Three-Dimensional Surface Motion with InSAR: Constraints from Multi-Geometry Data Fusion. Remote Sens. 2019, 11, 241. [CrossRef] 27. Zheng, M.; Deng, K.; Fan, H.; Du, S.; Zou, H. Monitoring and analysis of mining 3D time-series deformation based on multi-track SAR data. Int. J. Remote Sens. 2019, 40, 1409–1425. [CrossRef] 28. Wang, Z.; Zhang, R.; Wang, X.; Liu, G. Retrieving three-dimensional co-seismic deformation of the 2017 MW7. 3 Iraq earthquake by multi-sensor SAR images. Remote Sens. 2018, 10, 857. [CrossRef] 29. Zhao, Q.; Ma, G.; Wang, Q.; Yang, T.; Liu, M.; Gao, W.; Falabella, F.; Mastro, P.; Pepe, A. Generation of long-term InSAR ground displacement time-series through a novel multi-sensor data merging technique: The case study of the Shanghai coastal area. ISPRS J. Photogram. Remote Sens. 2019, 154, 10–27. [CrossRef] 30. Chen, X.; Zhang, Y. Impacts of urban surface characteristics on spatiotemporal pattern of land surface temperature in Kunming of China. Sustain Cities Soc. 2017, 32, 87–99. [CrossRef] 31. Fan, Y.H.; Li, W.C. Geological characteristics of the Pulang porphyry copper deposit, Yunnan. Geol. China 2006, 33, 352–362. 32. Zebker, H.A.; Villasenor, J. Decorrelation in interferometric radar echoes. IEEE Trans. Geosci. Remote Sens. 1992, 30, 950–959. [CrossRef] 33. Ferretti, A. Nonlinear subsidence rate estimation using permanent scatters in differential SAR interferometry. IEEE Trans. Geosci. Remote Sens. 2000, 38, 2202–2212. [CrossRef] 34. Baran, I.; Stewart, M.P.; Kampes, B.M.; Perski, Z.; Lilly, P. A modification to the Goldstein radar interferogram filter. IEEE Trans. Geosci. Remote Sens. 2003, 41, 2114–2118. [CrossRef] 35. Costantini, M. A novel phase unwrapping method based on network programming. IEEE Trans. Geosci. Remote Sens. 1998, 36, 813–821. [CrossRef] 36. Zhang, L.; Ding, X.; Lu, Z.; Jung, H.S.; Hu, J.; Feng, G. A novel multitemporal InSAR model for joint estimation of deformation rates and orbital errors. IEEE Trans. Geosci. Remote Sens. 2013, 52, 3529–3540. [CrossRef] 37. Schulz, K.; Boldt, M.; Even, M. Generalization of the CoVAmCoh analysis for the interpretation of arbitrary InSAR images. IEEE Int. Geosci. Remote Sens. Symp. 2012, 7444–7447. [CrossRef] 38. Zhu, W.; Zhang, Q.; Ding, X.; Zhao, C.; Yang, C.; Qu, F.; Qu, W. Landslide monitoring by combining of CR-InSAR and GPS techniques. Adv. Space Res. 2014, 53, 430–439. [CrossRef] 39. Meckel, T.A.; ten Brink, U.S.; Williams, S.J. Current subsidence rates due to compaction of Holocene sediments in southern Louisiana. Geophys. Res. Lett. 2006, 33.[CrossRef] 40. Bonì, R.; Bosino, A.; Meisina, C.; Novellino, A.; Bateson, L.; McCormack, H. A methodology to detect and characterize uplift phenomena in urban areas using Sentinel-1 data. Remote Sens. 2018, 10, 607. [CrossRef] 41. Tang, Y.Q.; Cui, Z.D.; Wang, J.X.; Lu, C.; Yan, X.X. Model test study of land subsidence caused by high-rise building group in Shanghai. Bull. Een. Geol. Environ. 2008, 67, 173–179. [CrossRef] 42. Franzius, J.N.; Potts, D.M.; Addenbrooke, T.I.; Burland, J.B. The influence of building weight on tunneling-induced ground and building deformation. Soils Found. 2004, 44, 25–38. [CrossRef] 43. Yu, B.; Liu, H.; Wu, J.; Hu, Y.; Zhang, L. Automated derivation of urban building density information using airborne LiDAR data and object-based method. Landsc. Urban Plan. 2010, 98, 210–219. [CrossRef] 44. Li, Y.; Zhang, N. Assessment of Groundwater Overdraft Zones in Kunming Basin. J. Yangtze River Sci. Res. Inst. 2017, 34, 35–38. Sensors 2019, 19, 4425 29 of 29

45. Holzer, T.L.; Johnson, A.I. Land subsidence caused by ground water withdrawal in urban areas. GeoJournal 1985, 11, 245–255. [CrossRef] 46. Woldai, T.; Oppliger, G.; Taranik, J. Monitoring dewatering induced subsidence and fault reactivation using interferometric synthetic aperture radar. Int. J. Remote Sens. 2009, 30, 1503–1519. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).