Ground Deformations Controlled by Hidden Faults: Multi


Download Ground Deformations Controlled by Hidden Faults: Multi


Preview text

remote sensing
Article
Ground Deformations Controlled by Hidden Faults: Multi-Frequency and Multitemporal InSAR Techniques for Urban Hazard Monitoring
Federica Murgia 1,*, Christian Bignami 2 , Carlo Alberto Brunori 2 , Cristiano Tolomei 2 and Luca Pizzimenti 2
1 Dipartimento di Ingegneria dell’Informazione, Elettronica e Telecomunicazioni (DIET), La Sapienza University of Rome, 00184 Rome, Italy
2 Istituto Nazionale di Geofisica e Vulcanologia, 00143 Roma, Italy; [email protected] (C.B.); [email protected] (C.A.B.); [email protected] (C.T.); [email protected] (L.P.)
* Correspondence: [email protected]
Received: 8 August 2019; Accepted: 23 September 2019; Published: 26 September 2019
Abstract: This work focuses on the study of land subsidence processes by means of multi-temporal and multi-frequency InSAR techniques. Specifically, we retrieve the long-term evolution (2003–2018) of the creeping phenomenon producing ground fissuring in the Ciudad Guzmán (Jalisco state, Mexico) urban area. The city is located on the northern side of the Volcan de Colima area, one of the most active Mexican volcanoes. On September 21 2012, Ciudad Guzmán was struck by ground fissures of about 1.5 km of length, causing the deformation of the roads and the propagation of fissures in adjacent buildings. The field surveys showed that fissures follow the escarpments produced during the central Mexico September 19 1985 Mw 8.1 earthquake. We extended the SAR (Synthetic Aperture Radar) interferometric monitoring starting with the multi-temporal analysis of ENVISAT and COSMO-SkyMed datasets, allowing the monitoring of the observed subsidence phenomena affecting the Mexican city. We processed a new stack of Sentinel-1 TOPSAR acquisition mode images along both descending and ascending paths and spanning the 2016–2018 temporal period. The resulting long-term trend observed by satellites, together with data from volcanic bulletin and in situ surveys, seems to suggest that the subsidence is due to the exploitation of the aquifers and that the spatial arrangement of ground deformation is controlled by the position of buried faults.
Keywords: subsidence; multi-temporal analysis; PS; SBAS; InSAR; urban monitoring; buried faults

1. Introduction
Ground subsidence is a geological phenomenon occurring in both uninhabited [1,2] and densely populated regions [3], or coastal plains [4,5]. Often, the subsidence of ground surfaces can be the result of the natural compaction of sediments or caused by anthropological activities like the extraction of groundwater, geothermal fluids, oil, gas, coal and other solids through mining [6]. Even if the hazard associated with subsidence is often different from the ones related to sudden events like earthquakes [7] because of their influence over a wide area, or like sinkholes formation, because of the limited extension of the affected area [8], the damages connected to surface slow sinking events are, however, extensive and with impacts across wide regions. Ciudad Guzmán (CG) is located in the Colima Volcanic Complex, near the active volcano Fuego de Colima [9] (Figure 1). It is a densely populated city characterized by moderate seismicity [10,11] and susceptible to intense withdrawal of water from aquifers widespread for civil and industrial purposes [12]. The already observed subsidence can be controlled by the presence of buried faults that can guide the generation of the

Remote Sens. 2019, 11, 2246; doi:10.3390/rs11192246

www.mdpi.com/journal/remotesensing

Remote Sens. 2019, 11, 2246

2 of 17

fissuring in the urban area [13]. The latter phenomenon produces a significant hazard that needs to be accurately assessed and monitored in order to prevent future impact on civil infrastructures. In order to analyze such surface movements, SAR Interferometry (InSAR) has proved to be a particularly useful tool capable of mapping ground deformation with a very high spatial resolution, with high accuracy [7,8,13]. In this study, in the framework of the time-series analysis of ground deformation, we applied two different interferometric techniques, i.e., Permanent Scatterers (PS), and Small BAseline Subset (SBAS) to the SAR dataset acquired by the European Space Agency (ESA) mission Sentinel-1. With this new set of ground deformation measurements, we extend the InSAR analysis of the CG ground deformations observed between 2003 and 2016 and described in [13,14], focusing on the most recent 2016–2018 period.
The final aims of this paper are to determine the correlation between the shapes of subsidence deforming and fracturing in Ciudad Guzmàn town with the position of the faults dislocating the bedrock under unconsolidated deposits filling the valley; to describe the ground subsidence evolution in urban areas by means the integration of SAR data acquired by different satellites platforms and sensors (Envisat, COSMO SkyMed, and Sentinel-1—in ENV, CSK and S1, respectively) and processed with multi-temporal InSAR techniques; to define a methodological approach for the study and the prevention of these hazardous geological processes.
2. Geological Overview
Ciudad Guzmán (1.500 m above sea level), with a population of about 105.000 inhabitants, is a Mexican city belonging to the Zapotlàn El Grande municipality in the Jalisco state. The town is located in the eastern side of a tectonic valley, the so-called Colima Graben [9]. The Colima Graben is the southern branch of the Colima-Tepic-Chapala triple junction located in the western sector of the Trans-Mexican Volcanic Belt (TMVB). The TMVB is a 1200 km long active continental volcanic arc originated by the subduction of the Cocos and Rivera plates along the Middle American Trench [15]. The TMVB is structurally divided into several regions, one of which is the Colima Graben (or Colima Rift). The Colima Graben is a structure that consists of three segments, the Northern Colima Graben (NCG), the Central Colima Graben (CCG) and the Southern Colima Graben (SCG). This region is considered the eastern limit of the Jalisco Block. The distinctive tectonic feature of this area is the presence of three structural patterns with NW, SW and EW orientation, associated with the local recent rifting processes [16,17].
Ciudad Guzmán (CG) is located in the South-Eastern part of the NCG and is surrounded by reliefs consisting of Late Miocene-Pliocene volcanic deposits, Jurassic-Eocene sedimentary and intrusive rocks [9]. The valley is filled by a sequence of quaternary lacustrine sediments, alluvium, colluvium and volcanic deposits of the nearby Colima Volcanic Complex (CVC), one of the most active volcanoes of the TMVB. The NCG (60 km long and 20 km wide) is flanked by parallel NNE-SSW-trending active faults dipping 70◦ toward the graben axis and is mainly characterized by a normal component of motion with respect to lateral strike-slip one.
The inhabited area of CG, as shown in Figure 1, is bound to the North by the wetland of Zapotlán basin and the sequence of sediments under CG urban area is mainly composed of weakly lithified volcaniclastic deposits characterized by a relatively low cohesion [10] as reported in Figure 2. This area is exposed to hazardous natural events such as landslides, volcanic eruptions and earthquakes [11], and it was hit by several seismic events in the past years (1911, 1931, 1932, 1941, 1975, 1985 and 2003) [18], mainly due to the tectonic activity of the volcanic arc and subduction zone [15–17]. In particular, the 1985 Mexico City earthquake (Mw 8.1) generated ground fissuring in the urban area of CG, even though the epicenter was located in Michoacán State, about 190 km far from CG. On 21 September 2012, without any correlation with seismic events, an alignment of soil fractures was formed in the city of CG. The orientation and position of the 2012 fractures are compatible with the ones opened during the 1985 earthquake.

Remote Sens. 2019, 11, 2246 Remote Sens. 2019, 11, x FOR PEER REVIEW
Remote Sens. 2019, 11, x FOR PEER REVIEW

3 of 17 3 of 18
3 of 18

FiguFrieg1u.reTTh1h.eeTlohloceaclatoitcoiaontnioofnoftohtfehtehstesutsdutyuddyayraearare.eaaT..hTTehhueepupupepprpelreerlfetlfeGtfGot ooGgoolgeolegElEaeartrEhtahrvvtiheiewwvileloowccaalloiizzceeasslittzhheees CCthooelliimmCaoalaianmndda
CaniuddCCaiiduuddGaauddzGGmuuàzznmmrààenngirroeenggsiioon(nisns((tiinhnetthhCeeoCCliomolliaimmGaarGGabrraeabnbeenanraeararee)aa)w) wiwthiitthhrerrseepsspepecectcttottootththheeeMMMiiiddddddllelee AAAmmmeeerrriiicccaaa ttrreenncchh (Cen(tCraeln-tWrael-sWteernsteMrnexMiicceoox))i.c.ToT)h.heTehppehhypyshisycisacilacmlalmampaapsphssohhwoowws tsshttehhseeussbuudbbdivdiviisviisioisoniononoffothtfhetehCCeooClliimomlaiamGGarraaGbbreeannb((eCCnGG())CiinGnt)thhiene threetethhrsreeeegemsseeeggnmmtsee,nnnttssa,,mnnaealmmy,eelltyyh,,etthhNeeoNNrtoohrretthrhneerrCnnoCClioomlliiammaGa GrGarrbaaebbneenn(N((NNCCCGGG),)),,CCCeenenntrttraralallCCCooollilimimmaaaGGGrrraaabbbeeennn(((CCCCCCGGG)))aanndd SoutShoeurntheCronliCmoalimGaraGberanb(eSnC(GSC) Gbe) tbweteweneetnhethJealJiaslcisocoanadndtthhteeheMMMiiccihhchoooaaaccácánnbbbllolooccckkksss.. CCiiuuddaadd GGuuzzmmàànnisis locatleodcaitnedthien SthoeuStho-uEthas-Etearsnteprnarptaorft tohfethNeCNGC,Ga,nadnidninthtehNe NorotrhtherenrnsisdideeooffttthhheeeNNNeeevvvaaaddoo dddeeeCCCooollliiimmmaaa((N(NNCCC))) aanndd aFFnuudeeggFooueddgeeoCCdooelliiCmmoaalim((FFaCC())FvvCoo)llvccaoannlciiaccnssittcrrusutccrttuuucrrteuessre((stthh(etehCCeooCllioimmlimaa VaVoVolloccalacnnainiccicCCCoomommpppllelexexx--C-CCVVVCCC)).)..

FiguFriegu2.reS2im. SpimlifipelidfieGdeoGleoogliocgailcaml ampapof oCf iuCdiuaddadGuGzumzmana.n. (V(VSS)) VVoolclcaanniicc SSeeddiiments;; ((BBTT)) BBrroowwnn TuffsT—ufPflsi—o-PQluioa-tQerunaaterryndareypodseitpso; s(iRtsb;)(Rebd) Rbedsb—edLsa—teLCartetaCcreeotuacse; o(Luis); L(Limi)eLsitmonees;to(Dnea;) (DDaac)itDesa—citEeas—rly FCirgeutaErceaerol2yu. sCS.riOemtnapctlehifoeiuebdso. OtGtonemothl,oethgbieocsattclohmmem,atpahteiocsfgcheCoeimluodagtaiiccdaglGecoruolozsmgs-iscaeancl.tciro(oVnsSsr)-esleVactoitvilocenatnoriectlhaeStievAdeAitm’opetrhnoetfisA;leA(iBn’TpF)riogBfuirlroeewi2nna (TmuoffdsF—ifigeuPdrleiof2r-oaQm(uma[ot1ed9r]inf)i.aerdyfrdoempo[1s9it]s);. (Rb) Red beds—Late Cretaceous; (Li) Limestone; (Da) Dacites— Early Cretaceous. On the bottom, the schematic geological cross-section relative to the AA’ profile in
Figure 2a (modified from [19]).

Remote Sens. 2019, 11, 2246

4 of 17

3. MateRremiaotlesSeanns. d201M9, 1e1t,hx oFOdRsPEER REVIEW

4 of 18

W3e. MpraotecreisaslsedantdwMoetSh1odSsAR stacks of 53 and 70 S1 images (TOPSAR mode) on ascending and descendingWpeaptrhosce(stsheed otwrboiSt1nSuAmRbsetarcskas roef 4539 aanndd7012S,1 riemsapgeecst(iTvOelPyS)AcRovmeordine)gona atsecmenpdoinrgalasnpdan from Novemdbeescren2d0i1n6g ptoathAs p(trhiel o2r0b1it8.numTwbeorsdairffe e4r9enantdm12u,ltrie-stpeemctpivoelrya)l ctoevcehrinnigqauetesm, pboortahl sipmanplferommented in SARScaNpoev©emsboefrtw20a1r6e (tSoaArmprailp2S0A18©. )T,wwoerdeiffaedreonpttemdufltoi-rtetmhepodraltatseecthsn:iqPuSes[,20b–o2th2]imanpdlemSBenAteSd[2in3–25]. As
far as thSAeRSSBcAapSe©msoeftthwoadre, (aSlalrtmhaepcSoA-r©e)g, wisetreereadoipntteedrfoerrothgerdamatassewtse: rPeSo[2b0t–a2i2n]eadndbySBaApSp[l2y3i–n2g5]a. Amsulti look factor ofafr4a×s th1einSBrAaSngmeetahnodd,aazllimtheuctho-rdeigriesctetrieodnsin,tleerafedrionggratmosaw1e5re×ob1t5aimnedpibxyealpspizlyeinognatmheulgtirloouonkd. Then, the intefarcfteorroogfr4a×m1sinwrearnegecoanmdpauzitmedutahnddireocntiloynst,hleeapdiinxgeltsosah1o5w× i1n5gmapcixoehl esirzeenocnetvhealguroeugnrde.aTtheernt,han 0.35
the interferograms were computed and only the pixels showing a coherence value greater than 0.35
were unwwerreaupnpwerdapfpoerdefaocrheapcahipr.air. We alsWoecaolnsosicdoenrseidderaedstaasctkacokfo9f 898CCSSKK iimmaaggeessspspananninnginagtiamteiminteerivnatlefrrvoaml OfrcotmobeOr 2c0to11betor 2011 to
SeptemSbeeprte2m01b5er[1240,1256][,1w4,h26ic],h wwheircehpwroecressperdocwesistehdthweitIhPTtAhemIuPTltAi-bmasuelltin-beatselcihneniqteuchen[2iq7u]eim[2p7l]emented in GAMimMplAem©enSteodftiwn GarAeMtMo Ain©tSeogfrtwataeretthoeintteemgraptoertahel taenmaployrsailsabnaeltywsies ebnetwEeNenVEaNnVdanSd1S(1F(iFgiugurree 3). The above-m3).enTthieonaebdovme-emthenotdioncoedmpmuettheosda sctoamcpkuotefsinatesrtfaecrkogorfamintse,rwferhoigcrhamarse, gwehnicehraatered gcoennesriadteedring only the SARcopnsaiidresricnhgaroanclytetrhizeeSdAbRypvaiarlsucehsaorafcstepraizteiadlbayndvatleumespoof rsaplabtiaalsealnindetsemlimpoirtaeldbawseitlihniens slipmeictiefidc ranges,
within specific ranges, aiming at minimizing the interferometric coherence loss. Then, the
aimingdaetfmorminaitmioinzitnimget-hseeriienstearnfderroesmideutarlictocpoohgerraepnhcicehloeisgsh.tsTahreene,stthimeadteedfobrymuastiniognthtiemSien-gsuelraier sVaanlude residual topogrDapechoimc phoesiigtihotns(aSrVeDe)sLtiemasat-tSeqduabryesuisnivnegrstiohne tSeicnhgniuqluaer [V25a]luaeppDlieedcotmo pthoesiintitoernfe(rSoVmDetr)icLestaasctk-Squares
inversi[o1n3]t.echnique [25] applied to the interferometric stack [13].

FigureF3i.guMrea3p.sMoafptshoef dtheefdoerfmoramtiaotnionrartaeteffrroomm tthheeEENNVVdadtaasteats(e2t00(23–020031–02t0im10e ptiemrieodpdeersicoednddiensgcaenndding and ascendainscgenpdaitnhgs,puatphps,eurpppaenr eplasnoenls tohnethleeftleaftnadnrdigrihgth,tr, ersepspeeccttiivveellyy))aannddfrforommthtehCe SCKSdKatdaasetat s(2e0t1(12–011–2015 time pe2r0i1o5dtimasecpeenrdioidngaspceantdhi,nlgopwatehr, ploawneerlp).anBeol)t.hBoththetEheNEVNVanadndCCSSKK ddaattaasseetstswwereerperopcreossceedssuesdingusing the
the IPTA multi-baseline technique. The white circles indicate three selected sites individuating
IPTA mduiffletir-ebnat sdeislipnlaecetemcehnntsiqrautee.s:T(Ah)eswtabhlieteorcliirgchlet-spoinsidtiivceagterotuhnrdeevesleolceitcietesd(toswitearsdinthdeisvaitdeluliatet)i;n(gB)different displacements rates: (A) stable or light-positive ground velocities (toward the satellite); (B) intermediate
values of negative velocities (moving away from the satellite), near the fractured area; (C) highest
values of negative ground velocities. A ground urban survey performed in CG in November 2012
assessed the deformations and associated fractures (black points alignment on the maps) that occurred
on 21 September 2012 [13].

Remote Sens. 2019, 11, x FOR PEER REVIEW

6 of 18

Remote Sens. 2019, 11, 2246

5 of 17

As already mentioned, ground deformation time series of the S1 dataset were evaluated by

means of both the SBAS and PS techniques. The obtained LOS (Line of Sight) mean ground velocity

3m.1.apSsen(tFiingeul-r1eS3B)AhSigPhrloigcehstsitnhge presence of a subsidence (i.e., the distance satellite-target increases) in

the NInWordareeratootfrCacGk,tihneaNgoreveemmebnert 2w0i1t6h–tAhperriel s2u01lt8s tsehmopwonrailne[v1o3l]u.tTiohnisosfugbrosuidnidngsuabrseiadiesnbceoradffeercetdingto CthGe, EwaestabpyptliheedlothcaetSiomnaolfl tBhAe s2e0l1in2esuSrufbacseetfi(sSsBuAreSs,. [C2r3o–s2s5i]n–gFtihgue rreup4atu) rteecahlinginqmueentot, pwreofcoeussndthaesttiambele saerreieas(oSfouSt1h-dEaatsatseertn. sTidhee oSBf ACSG)a,lgasoroitbhsmervisedbainse[d13o]nwainthaEpNprVoparniadteCcSoKmdbeinfoartmioantioofndtiiffmeeresnetriiaels in(Fteigrfuerreog2r)a. mAsspirnod[1u3c]e,dwbey cdoantasipdaeirresdchtharreaectesritiezse,d(bAy; aBsmanadll sCp,aFtiiaglu(rpeer3p)etnodiacnualalyrz) ebatsheelingeroaunndd lidmefitoerdmtaetmiopnobreahl asvepioarraintiodniffienreonrdt epratrotsloimf tiht ethceitsyp. aStpioe-ctiefimcaplolyr,apl odiencto(rAre)laistilooncaptehdeninomtheenSaE. Apat rtthoef bGegCi,nwnihnegreotfhtehdeiSsAplRacdemataenptrsorcaensgseinfgro, mwe0steolselcitgehdtlay mpoasxiitmivue mvalsupeastiianltbhaessealitnelelietequLaOlSto(i.1e5.,0dmistaanndce asmatealxliitme-utamrgteetmrepdouracledb)a;sseiltien(eBe)qisulaolctaote7d2 ndeaayrstahnedNroerstuhletirnng2i0n1216fi0ssausrceenadliginngmaenndt o2b08sedrvesecdednduriningg ptahiersf.ieTlhdesuSBrvAeSy mofeNthoovdecmobmepr u2t0e1s2d[e1f3o]r, manadtiosinteti(mC)e isseirnietshaenNdWrepsiadrut aolfttohpeocgitrya,pwhihcehreeitghhetssuubsisnidgetnhcee Spinhgeunloamr Veanloune iDsemcoomrepposriotnioonu(nScVeDd.) ALesaistt-cSaqnubaerensointevde,rseivoenntethchonuigqhueth[2e3t,w24o].dIet fiosrwmoarttihonsaryaintegmthaapt,s inreotrrideevretothoebstaaimnea scpleaatniaelr afirnraalndgiesmpleancteminentetrmmaspo,fafdteerfotrhme afitrisotntipmaet-tseerrnise,stehsetimvaaltuioens ,awree napotpleiqeduaal, ceussptoemciaalltymionspthheerhiicgfihl-treartiengantod tlheessstcaochkeroefnutnpwarrat popf ethdeinctietyrfe(ir.oe.g,rtahme sn.oTrhthe-wdiemsteenrsnioonnseo) fdtuhee tuosethde fiflatcetr tahraet1t2h0e0twmo×te3c6h5ndiqauyse,sraerperedsieffnetraetnivt.e of a low pass filter in space and a high pass filter in time.

FFigiguurere44. .DDeeffoorrmmaattioionnrraatteemmaappssffrroommththee22001166––22001188SS11ddaatatasseett(a(asscceennddininggppaathth).).PPSS((aa))aannddSSBBAASS((bb)) tetecchhnniqiquueessreressuultlsts. .TThheewwhhititeeccirircclelessininddicicaateteththeesseeleleccteteddssitietessininddivivididuuaatitninggddiffifefererennttddisipsplalacceemmeenntsts raratetes:s:(A(A) s)tasbtaleb;l(eB;) (iBn)terinmteedrmiaeted/ifaratec/tfurraecdtu; (rCed) ;hi(gCh)eshtivgahleusets.vTahlueebsl.acTkheellibplsaecsksuerllriopusneds asuprarrotiucnudlara apreaartoicfuClaGr tahraetawofilCl bGe tdhiastcuwsislel dbeindtishceuDssiesdcuisnsitohne sDeicstciounss. ion section.
4.1. CAosmaplraeraisdoyn mbeetwnteieonnSedB,AtSheanadboPvSePdreo©sccesrsibinegd multi-temporal processing was done with the SBAS technique implemented into the Sarscape software. As it is well known, the phase unwrapping step, foreseWenitohntlhyeinaimtheoSf BcrAoSsst-evcahlindiaqtuineganthdenroetsuinltsthaechPiSevoende,urseinpgordtsiftfheree2nπt tmecohdnuiqleueinateprpfelireodgrtoamthse tosaamceoanrteinauaonudstphheassaemreepSrAeRsednatatatisoent,,waceccoordmipnagrteoththeeremsuinltims oizf aPtSioanndofSaBpArSopteecrhcnoiqstufeusn. TcthioenP.STahned aSlgBoArSithremtrisetvaertds vtoeluocniwtiersappathtterpnasirasrefrvoemryasipmoilnatr., Talhtheosuog-chaltlheedyrsehfeorwenscoempeodinistcirseapsasnucmieesdinstaerbmles doufrdinegfotrhme apteiorinodraotef ivnatelurest.mTahinislyimdpuleemtoetnhteattiwonoddoieffsenreonttpiemrmplietmtoenmtatniuonalsl.yTsheilsecitmthpelierseftehraetncthe,e bpurtoacpepdluierse aorfes-flelaetctitninggofththeecionhterrfeenrtompoeitnritcssatancdk,ibnapseadrtoicnutlhaer stehleecctoionnnoecftsiomn egcroapnthroolfptohinetps ainirtshteo uenvwalruaaptpeeadrednaotatstehte. Tsahmises,tpeposcsaibnlylelaedadtoinugntaoccaobuiansteidn bthiaesdienfothrme fiantiaolnprraotdesucvtasl,uief st.hAe crteu-flalaltyt,eSnBinAgS dkideenpostiwn oarckcofiunnet. dTioscorveeterctoamrgeetshiasnidsstuhee, wbaecskcsaclaetdtetrheeddrefsopromnasetioton trhaetermadaapr aenchdothise ftoimrmeesderbieys tohfe dseigfonraml bataicoknscaactctoerdedinfgrotomthalel pthaerttaorfgtehtes cpirtyes(ein.et.,inthtoe tshoeutrhes-eoalustteiorn pcealrlt. IonfsCteGa)dt,hthatesPhSomwedthaodstoalbolgey bceohnasvidioerrsinonthlye wthheorleespdoantasseeto,vienrcltuhdeitnimg tehienPteSrvoanlef.oMr eoarcehpsriencgilseelsyc,awtterseerl,eacltseodaat garsouubppoixfepl osicnatlse. cTohnetadinifefderiennacecirncutelarmr asroeaf sopfa2t5ia0lmcodviearmageeteirs, dchuaeratocttehreizfeadctbtyhtahteSsBaAmSe, acshaarlarecatedryisstitcasteidn,tkeremepssoifn toapccoogurnapt hfoyr, adnisdtraibvuetreydssimcaitltaerrderesf.oIrnmdaeteiodn, tthrenidnt,earnfedrowgerascmasleadrethme uenlttii-rleoopkroeduacntdbysptahteiaalvlyerfailgteroefd thweitdhefGoromldasttieoin rnaoteiseanrdedturecntidosnrefitlrtieerve(dthiant tihsataamrean. datory step for this kind of multi-temporal processing); therefore, the coherence increases in SBAS, and consequently the spatial coverage 3i.n2c. rSeeansteins eals-1wPeSll.PTrohciessissinpgarticularly true in the area where fast subsidence occurs: over these regions,
the PWSemalestohpordo,cweshseicdhthdeoseasmneotSu1 ndwatraaspettwheithinttheerfmeruolgtir-atemmsp, oisraulnmabetlheotdoo“lfooglylo, wna”mseulcyh, thtreePnedr,stihstuesntit Slcoastetsereprosss(PibSle[2p0e–r2s2i]s–teFnigtusreca4ttbe)reirms pcleamndeindtaetdesi.nInSApRaSrctiacpuel©ar,sothftewarreetr. iePveSdisPaSn mopapposrtsuhnoiwstica deformation rate up to 15 mm/year in correspondence of the fractures, and a faster subsidence

Remote Sens. 2019, 11, 2246

6 of 17

deformation measurement method, i.e., it is able to measure deformation only over the available PSs. Usually, the PS density is low in vegetated, forested and low-reflectivity areas (e.g., very smooth surfaces), and in steep terrain facing the radar sensors [20]. By contrast, PSs are usually abundant on buildings, or urban areas in general, so this multi-temporal technique represents a suitable tool for our study. As it will be discussed later, the PS algorithm implemented in the above mentioned software does not perform the unwrapping, meaning that it algorithm avoids to analyze the phase, since at this stage it may contain unknown signal contributions. Therefore, the degree of “reliability” of each single pixel is evaluated looking at the amplitude dispersion index (i.e., the ratio between the temporal standard deviation of the amplitude and the temporal average of the amplitude), meaning that a pixel characterized by a similar amplitude during all acquisitions is expected to have a small phase dispersion [28].
As usual, in order to reduce the atmospheric disturbances, a spatial and temporal filter is applied. The parameters we selected are the same set for SBAS elaboration.
4. Results
As already mentioned, ground deformation time series of the S1 dataset were evaluated by means of both the SBAS and PS techniques. The obtained LOS (Line of Sight) mean ground velocity maps (Figure 3) highlight the presence of a subsidence (i.e., the distance satellite-target increases) in the NW area of CG, in agreement with the results shown in [13]. This subsiding area is bordered to the East by the location of the 2012 surface fissures. Crossing the rupture alignment, we found a stable area (South-Eastern side of CG), as observed in [13] with ENV and CSK deformation time series (Figure 2). As in [13], we considered three sites, (A; B and C, Figure 3) to analyze the ground deformation behavior in different parts of the city. Specifically, point (A) is located in the SE part of GC, where the displacements range from 0 to slightly positive values in the satellite LOS (i.e., distance satellite-target reduced); site (B) is located near the Northern 2012 fissure alignment observed during the field survey of November 2012 [13], and site (C) is in the NW part of the city, where the subsidence phenomenon is more pronounced. As it can be noted, even though the two deformation rate maps retrieve the same spatial arrangement in terms of deformation patterns, the values are not equal, especially in the high-rate and less coherent part of the city (i.e., the north-western one) due to the fact that the two techniques are different.
4.1. Comparison between SBAS and PS Processing
With the aim of cross-validating the results achieved using different technique applied to the same area and the same SAR dataset, we compare the results of PS and SBAS techniques. The PS and SBAS retrieved velocities patterns are very similar, although they show some discrepancies in terms of deformation rate values mainly due to the two different implementations. This implies that the procedure of selecting the coherent points and in particular the connection graph of the pairs to evaluate are not the same, possibly leading to a bias in the deformation rates values. Actually, SBAS keeps in account discrete targets and the backscattered response to the radar echo is formed by the signal backscattered from all the targets present into the resolution cell. Instead, the PS methodology considers only the response over the time interval for each single scatterer, also at a subpixel scale. The difference in terms of spatial coverage is due to the fact that SBAS, as already stated, keeps in account for distributed scatterers. Indeed, the interferograms are multi-looked and spatially filtered with Goldstein noise reduction filter (that is a mandatory step for this kind of multi-temporal processing); therefore, the coherence increases in SBAS, and consequently the spatial coverage increases as well. This is particularly true in the area where fast subsidence occurs: over these regions, the PS method, which does not unwrap the interferograms, is unable to “follow” such trend, thus it loses possible persistent scatterers candidates. In particular, the retrieved PS maps show a deformation rate up to 15 mm/year in correspondence of the fractures, and a faster subsidence reaching 40 mm/year in the NW of CG where the velocity retrieved from the SBAS processing reaches 45 mm/year. The SBAS and PS

reRpemoortteeSdentsh. 2e0v19e,l1o1c, ixtiFeOsRoPfEaElRl tRhEeVsIEcaWtterers included in a circular area of 250 m of radius surroun7doinf g18 each of the three considered sites (named A, B and C) and highlighting three different displacement rareteasc.hing 40 mm/year in the NW of CG where the velocity retrieved from the SBAS processing reaches 45 mTmhe/ysecaart.teTrhpeloStBsAhSigahnlidghPtS dreifsfuerltesnwt ebriaesceosmapnadreddis(pFeigrsuiroens5)d. IenpethnedisncgattoenrptlhoetsaonfaFlyigzuedre a6reaare RcerhmeaoprteoaScretnesdr.i2zt0he1de9,vb11ey,l2od2ci4if6tfieersenotf dalelftohremsactaitotnerreartseisn. cluded in a circular area of 250 m of radius surrou7nodf i1n7g each of the three considered sites (named A, B and C) and highlighting three different displacement rates. resultsTwheerseccaottmerppaloretsd h(Figighulirgeh5t).dIinffethreenstcabtitaesrepsloatnsdofdFiisgpuerresi6onasredreeppeonrtdeidngthoenvetlhoeciatineasloyfzeadll tahreea sccahtaterraecrtserinizceldudbeyddiniffaerceirnctudlaerfoarrmeaaotifo2n5r0amteso. f radius surrounding each of the three considered sites (named A, B and C) and highlighting three different displacement rates.
Figure 5. Mean LOS ground velocity of S1 ascending stack: the SBAS vs. PS cross-comparison analysis for the A, B and C selected sites of Figure 4.
In Table 1 statistics relative to the comparison of the two multi-temporal techniques adopted in this wFiFogiurgkruerfe5o.r5M.tMheeaenathnLrLOeOeSSgscgroaruottunednrdpvlevoleotlscoicotyiftyoFfoigSf 1uS1raesacs6ecneadnridenigsnugsmtsatcmakc:akt:rhitzheeeSdBS.BAASSvsv.s.PPSScrcorosss-sc-ocommpparairsiosnonanaanlaylsyissis
for the A, B and C selected sites of Figure 4. for the A, B and C selected sites of Figure 4.
In Table 1 statistics relative to the comparison of the two multi-temporal techniques adopted in this work for the three scatterplots of Figure 6 are summarized.
FFiigguurree66.. LLOOSS ggrroouunndd vveelloocciittyyoofftthheeSS11aasscceennddininggssttaacckk: : tthheeSSBBAASSvvss.. PPSSccrroossss--ccoommppaarriissoonnaannaallyyssiiss ffoorrtthheeAA,,BBaannddCCsseelleecctteeddssiitteessooffFFiigguurree44..
TThhee sdceaftotremrpaltoiotsnhriagtehsliogfhtthdeiAffesrietentshboiawseashainghd sdtaisnpdearrsdiodnesvdiaetpioennddiinffgeroennctehief caonmalpyazreedd awrietah cthhaeriracdFtyeigrnuiazrmeed6ic.b.LyTOdhSiiffsgeriosreubnnetdcdaveueflsooecrimtthyaeotiSfoBtnhAerSaSt1aepass.pcreonadcinhgpsrtoacvki:dtehseaSBsmASovosth. PeSr scroolsust-icoonmwpairthisornesapneaclyt stios the PS oInnfeo,TraatnbhdleeAt1h, isBstaaistnidsmtCiocsrseerleeelcvatietdidvesenittteoisntohafeFsictgaoubmrleep4aa.rriesao.nTohfetrheeastwonommauyltbi-etefmoupnodrailntethchenfaiqcut ethsaatdtohpetSedBAinS tahligsowriothrkmfoarpthpeliethsretehescaGttoelrdpsltoetisnoffiFltiegrur[e269]arteosummumltai-rliozoekde. d pairs, thus generating smoother interTfeThrehoegdrdeafeomfrosmr.mEatavitoeinnontrhartoaeutsegsohfotfthhtehe esAuAbsistiedtesehnshocwoe wpaaahtthiegirghnhsistsatnnaendadarraldyrdidddeevenvitaitaciatoilonfnodrdiffbifeofretehrnemcneceeitfhicfoocdmosm,pipanraetrhdeedwawirteihtah twhtehierirrdedythynenaamsumibcis.ci.dTTehhniicsseiissphbbeenccaoauumsseenttohhneeiSsSBBrAeAlSeSvaaapnpptpr,rotohaaecchPhSpprmoroevvtihdidoedsesaseasemsmmosoottoohtoehrveesrroseloustltiuimotinaotnwe tiwthhietrhdeesrfpeoserpcmteactottiottohne trhaPetSePocSnooem,napen,adraentddhiwsthiitsshmistohmreeoSerBveAiedSveindrteesinnutlati.nsHtaaobswlteaebavlreeeraa,.rteThahe.emTrheaaexsiromenausmmoanydmebvaeiyafotbiuoenfdobuienntwdtheienenftahcteht etfhatacwtt tothheamtSutBhltAei-S SteBamlAgpSoorairtlahglmotreictahhpnmpiqlaiuepespsltiisheesletsGhseothlGdasontle9disnmtemifni/ltyfierrl.te[r29[2]9]totommuultlit-il-oloookkeedd ppaaiirrss,, tthhuuss ggeenneerraatitninggsmsmoootohtehrer initnetrefrefreorgorgarmams.sE. Evvenenththoouugghhththeesusubbsisdidenenceceppatattetrenrnisisnneaeralrylyididenentitciaclaflofrorbobtohthmmetehthodods,si,ninthtehearaeraea wwhehreerethtehesusbusbisdiednecnecephpehneonmomeneonnonisirserleelveavnatn, tt,htehePSPSmmetehtohdodseseemems stotoovovereersetsitmimataetethteheddefeofromrmataitoinon rartaetecocmopmapreadrewditwhitthhe tShBeASSBreAsSulrt.esHuoltw. eHvoewr, tehveemr, atxhiemmumaxdimevuiamtiodnebveiatwtioenenbthetewtweeonmtuhleti-ttwemo pmouraltlitetcehmnpiqouraesl tiesclhenssiqthuaens i9s mlesms/tyhra. n 9 mm/yr.

Remote Sens. 2019, 11, 2246

8 of 17

Table 1. Cross validation of the two multi-temporal outputs used for the S1 data processing methods. The values are relative to the total component of the deformation rate in the LOS of S1 on the ascending path. The same comparison could not be made for the descending path because of the lack of CSK images on that orbit.

#

Bias [mm/y]

Std Difference [mm/y]

Correlation

A

0.832

0.731

0.767

B

2.903

2.794

0.975

C

8.866

2.587

0.957

4.2. Deformation Time-Series Analysis
Aiming at completing the multi-temporal InSAR study carried out by means of the ENV [13] and CSK [14,26] datasets, we added the information retrieved by exploiting the S1 acquisitions along the ascending path only, processed with the SBAS algorithm. For this aim, we first had to fill the gaps in between the three time-series obtained from the different sensors. The problem here was to find the way to link the series, since the three datasets are time-gapped [30]. For the sake of clarity, it is worth saying that the analysis described hereafter refers to the three control points (i.e., A, B, and C) identified and presented above, but the procedure can be applied to the whole map if there is a consistent number of high coherence pixels in common between the different time-series. We first assumed a linear trend for the deformation (Figure 4) since the dynamic of the subsidence resembles a linear evolution with the time. In this way, the gaps between each time series (ENV, CSK, and S1) were filled-in through a linear fitting and by extrapolating the deformation slope. In particular, we performed a linear forward prediction of the previous time series up to the time of the first acquisition of the following dataset (e.g., the ground displacement of the ENV time-series has been prolonged until the first acquisition time of the CSK one). Then, the following time-series were scaled to a value given by the corresponding one assumed by the extrapolation of the previous deformation series in correspondence with the time of the first acquisition date of the considered dataset. The scaling step is fundamental since each time series was evaluated independently and so each long-term ground subsidence observations starts from 0 mm displacement.
Clearly, it is not a rigorous procedure, but the lack of ground truth information and of a geological/hydrological dynamic model did not permit us to apply another approach. However, the obtained results, including the changes in the slope of the displacement, are in agreement with the annual reports of the hydrological situation in the administrative region to which CG belongs [31] and with the agreement for the sustainability of the area [32] that will be presented in the Discussion section. It is worth noticing that in order to make the results comparable, we re-projected the S1 and ENV data on the CSK ascending LOS (LOSCSK). This implies the assumption that the deformation is only vertical, neglecting the east-west and the north-south components of the deformation [13,14]. Actually, due to the lack of CSK data along the descending path CSK, the East-West component is not retrievable for the overall monitoring period (2003–2018). We can also add that as regards the North-South direction, it is well known that quasi-polar orbit of space-borne SAR has a very weak sensitivity with respect to deformation. The procedure applied to the ENV and S1 time-series refers to the geometry sketched in Figure 7, where the situations of the CSK and S1 satellite are reported.
It is worth noting that the same procedure is applicable to the ENV dataset considering its standard incidence angle (i.e., 23◦). Looking at the acquisition geometry, we can see that:

S = X − D = Ccosδ − D,

(1)

Being δ the angle of the difference of the incidence angles of S1 and CSK satellites. Moreover,

D = ycosδ = Ltgϕcosδ = Ltgδcosδ,

(2)

Remote Sens. 2019, 11, 2246

9 of 17

and since L = Ctgθ, the resulting value for S is given by the following relationship:

S = Ccosδ(1 − tgθtgδ),

(3)

Figure 8 shows the results for the three considered sites. As it can be easily noted, the retrieved

trend from S1 data is very similar to the ones observed from the ENV and CSK datasets. It is worth

nRoetmhointegSetnhsa. 2t0s1u9,b1s1i,dxeFnOcRe PinEEtRheRElaVsIEt Wtwo years seems to drop down.

9 of 18

Figure 7. Acquisition geometry of the S1 and CSK satellites used to project the vertical component of deformation of the S1 dataset onto the CSK line of sight. It is worth remarking that even though the ENV LOS is not present in this figure, the same approach has been applied to such dataset considering the proper incidence angle, i.e., equal to the standard acquisition off-nadir of 23°.

It is worth noting that the same procedure is applicable to the ENV dataset considering its standard incidence angle (i.e., 23°). Looking at the acquisition geometry, we can see that:

𝑆 𝑋 𝐷 𝐶𝑐𝑜𝑠𝛿 𝐷,

(1)

Being δ the angle of the difference of the incidence angles of S1 and CSK satellites. Moreover,

𝐷 𝑦𝑐𝑜𝑠𝛿 𝐿𝑡𝑔𝜑𝑐𝑜𝑠𝛿 𝐿𝑡𝑔𝛿𝑐𝑜𝑠𝛿,

(2)

and since L=Ctgθ ,the resulting value for S is given by the following relationship: Figure 7. Acquisition geometry of the𝑆S1 an𝐶d𝑐𝑜C𝑠S𝛿K1sate𝑡ll𝑔it𝜃es𝑡𝑔u𝛿sed, to project the vertical component of (3)
Figure 7. Acquisition geometry of the S1 and CSK satellites used to project the vertical component of
dFdeigfeoufrormrema8taitsoihnoonowfoftshttehheSe1Sr1deadstuaatlsatesstefotonortnotthotehtehtheCrCSeKSeKlcionlinenseoifodsfeisgriehgdth.tsI.tiItitesissw. wAorostrhitthrceramenmarbakreikneignagsthitlahytanet voeevtneednth,tohthuoeguhgrhethttrehieeved trendENEfrNVoVmLOLSOS1Sisdisnanotatotpisprervseeesnreytnitnsiintmhthiisliasfirfgitguourert,eht,hetheoesnasemasmeoeabpasppeprrorvaoecadhchhfrhaosamsbbeeethneneapaEppNplileVidedatontodsusCuchcShdKdatadatasaetsatestceoctonsns.isdIitdereiisrniwgngorth
the proper incidence angle, i.e., equal to the standard acquisition off-nadir of 23◦.
nothintghethparotpseurbisnicdideenncceeianntghlee, lia.es.t, etqwuoalyteoatrhsessetaenmdsartdo adcrqoupisditoiownno.ff-nadir of 23°.

It is worth noting that the same procedure is applicable to the ENV dataset considering its standard incidence angle (i.e., 23°). Looking at the acquisition geometry, we can see that:

𝑆 𝑋 𝐷 𝐶𝑐𝑜𝑠𝛿 𝐷,

(1)

Being δ the angle of the difference of the incidence angles of S1 and CSK satellites. Moreover,

𝐷 𝑦𝑐𝑜𝑠𝛿 𝐿𝑡𝑔𝜑𝑐𝑜𝑠𝛿 𝐿𝑡𝑔𝛿𝑐𝑜𝑠𝛿,

(2)

and since L=Ctgθ ,the resulting value for S is given by the following relationship:

𝑆 𝐶𝑐𝑜𝑠𝛿 1 𝑡𝑔𝜃𝑡𝑔𝛿 ,

(3)

Figure 8 shows the results for the three considered sites. As it can be easily noted, the retrieved trend from S1 data is very similar to the ones observed from the ENV and CSK datasets. It is worth nothing that subsidence in the last two years seems to drop down.

Figure 88. .ExEaxmamplpeloef tohfretehrdeeefodrmefaotrimonatrioendtsrienntdhse (iAn),t(hBe) a(And),(C(B) )sitaensdof (FCig)usriete4s. TohfeFdiigsuprlaece4m. eTnhtes wdiesrpelaescteimeantetds bwyeEreNeVs,tCimSKateadndbSy1EdNatVa,seCtsSK(fraonmdaSsc1ednadtiansgeotsrb(firtoomnlya)s.cTehnediEnNgVorabnidt So1nldyi)s.pTlahceemENenVt data are projected along the LOSCSK. Each of the three colors (black, blue and red) represents the three satellite-derived displacement measures. The A displacement points are relative to the part of the CG that lies on the bedrock and thus shows a stable behavior. The B displacement points are relative to a site located in correspondence of the cracks that opened on September 2012 (indicated by the green line). The C displacement points are relative to the northwestern part of CG, with high deformation rates.

displacement points are relative to a site located in correspondence of the cracks that opened on September 2012 (indicated by the green line). The C displacement points are relative to the northwestern part of CG, with high deformation rates.

4.3. Vertical and East-West Components Estimation
Remote Sens. 2019, 11, 2246

10 of 17

The vertical (Figure 9) and East-West (Figure 10) deformation rate components were calculated

combining the ascending and descending S1 data following the well-known relationships [33] as 4fo.3ll.oVwerst:ical and East-West Components Estimation

combTinhienvgethrteicaaslc(eFnidgiunrgea9n)dadnedsEcean𝑣sdt𝑈-i𝑃WngesS𝑑t1𝑎(d𝑠F𝑐ai𝑠gt𝑖au𝑛fr𝜃oe𝑑l𝑒l1𝑠o0𝑐w) idn𝑑egf𝑑o𝑒th𝑠r𝑐me𝑠𝑖wa𝑛te𝜃ilo𝑎l𝑠-n𝑐krnaotwe ncormelaptoionnesnhtispws [e3r3e]caaslfcoulllaotweds: 𝑠𝑖𝑛 𝜃𝑑𝑒𝑠𝑐 𝜃𝑎𝑠𝑐

𝑣

vUP = 𝑑da𝑑sc𝑒s𝑠i𝑐n𝑠θ𝑖d𝑛esc𝜃−𝑎d𝑠d𝑐escsin𝑑θ𝑎a𝑠sc𝑐𝑠𝑖𝑛𝜃𝑑𝑒𝑠𝑐
sin(θ +θasc)

𝐸𝐴𝑆𝑇 𝑊𝐸𝑆𝑇
v−

= d𝑠de𝑖s𝑛cdseisn𝜃cθ𝑑a𝑒sc𝑠−𝑐dascs𝜃in𝑎θ𝑠𝑐desc

EAST WEST

sin(θdesc +θasc )

(4) (4)

FFiigguurree 99.. VVeerrttiiccaall ddeeffoorrmmaattiioonn rraattee rreettrriieevveedd ffrroomm SS11 aasscceennddiinngg aanndd ddeesscceennddiinngg ddaattaa.. IInn tthhee AA--AA’’ Remotvve eeSlleoonccs.iitt2yy01pp9rr,oo1fif1i,llexe,,FttOhhRee PhhEooErriiRzzooRnnEttVaaIllEaaWxxiiss ggiivveess tthhee ddiissttaanncceess ((mm)),, tthhee vveerrttiiccaall aaxxiiss tthhee vveelloocciittiieess ((mmmm//yy1rr)1):: of 18
ppoossiittiivvee aanndd nneeggaattiivvee vvaalluueess iinnddiiccaattee EEaassttwwaarrddaannddWWeessttwwaarrddmmoovveemmeenntstsrreessppeecctitviveelyly. .
Figure 10. East-West deformation rate retrieved from S1 ascending and descending data. In the A-A’ vFeilgoucritey1p0r.oEfialset,-Wtheeshtodreizfoornmtaaltaioxins rgaitveersetthreiedviesdtafnrocems (Sm1)a,stcheenvdeirntgicaanl adxdisetshceenvdeilnogcitdieasta(.mInmt/hyer)A: t-hAe’ nveegloactiivtyepvreoloficleit,ythvealhuoersizidonentatlifayxtihs egisvuebsstihdeendciestaarnecae.s (m), the vertical axis the velocities (mm/yr): the negative velocity values identify the subsidence area.
As expected, the movement on the North-Western side of CG is mainly vertical with a small horizontal component of deformation. The maximum vertical deformation rate reached on the velocity profile A-A’ is about −60 mm/year, while on the same profile, the E–W deformation rate

Preparing to load PDF file. please wait...

0 of 0
100%
Ground Deformations Controlled by Hidden Faults: Multi