ISSN: 2476-6909; ECOPERSIA. 2020;8(2):77-87

C I T A T I O N L I N K S

Copyright© 2019, TMU Press. This open-access article is published under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License which permits Share (copy and redistribute the material in any medium or format) and Adapt (remix, transform, and build upon the material) under the Attribution-NonCommercial terms.

Land Subsidence Modelling Using Particle Swarm Optimization Algorithm and Differential Interferometry Synthetic Aperture Radar

[1] Assessment of climate change impacts on groundwater recharge for different soil ... [2]Spatial and temporal analysis of monthly stream flow deficit intensity in ... [3] Landsubsidence in Iran caused by widespread ... [4] Assessment of citrus water footprintcomponents ... [5] Assessment of hydro-meteorological drought ... [6] Characteristics andtrends of land subsidence in ... [7] Groundwater ... [8] Spatial prediction models for shallow ... [9] Detection of land subsidence in Semarang ... [10] Modeling alluvial aquifer usingPMWIN ... [11] Watershed health characterization using reliability-resilience-vulnerabilityconceptual framework based on hydrological ... [12] Detection of land subsidence inKathmandu ... [13] Land subsidence: What is it and why is it ... [14] Prediction of groundsubsidence in Samcheok City... [15] Assessment of ground subsidence using GIS and ... [16] Land subsidence risk assessment case ... [17] Investigation into subsidence hazards due to... [18] Assess the potential of land subsidence ... [19] Introducing a new framework formapping ... [20] Comparison of vulnerability of the southwest Tehran ... [21] Optimization of the ALPRIFT method using a support ... [22] A new approach to determine probable ... [23]Groundwater ... [24] Review: Regional land subsidence accompanying ... [25] Delimitation of ground failure zones due ... [26] Principles of applied ... [27] Groundwater vulnerability map ... [28] Enviromental ... [29] Particle swarm ... [30] Application of fuzzy particle swarmoptimization ... [31] Optimization of water allocation during ... [32] Path planning for aplanar hyper-redundant manipulator ... [33] Modeling and optimizing lapping process ...[34] A hybrid artificial neural network and ... [35] Topographic mapping from interferometric ... [36] Detection of land subsidence due to excessive ... [37] Dinsar-Based detection of landsubsidence and ... [38] Land subsidence modelling using tree-based machine learning ... [39]Tehran-Shahriar Plain subsidence due to excess extraction of underground ... [40] Monitoring its land subsidence and its relation to groundwater harvesting ...

Aims Land subsidence is one of the phenomena that has been abundantly observed in Iran’s fertile plains in recent decades. If it is not properly managed, it will cause irreparable damages. So, regarding the frequency of subsidence phenomenon, the evaluation of the potential of the country’s fertile plains is necessary. Towards this, the present study is formulated to assess the vulnerability of the Tehran-Karaj-Shahriyar Aquifer to land subsidence.Materials & Methods The vulnerability of Tehran-Karaj-Shahriyar Aquifer was determined using the GARDLIF method in a Geographic Information System (GIS) environment. Seven parameters affecting ground subsidence including groundwater loss, aquifer media, recharge, discharge, land use, aquifer layer thickness, and the fault distance were used to identify areas susceptible to land subsidence. Then, they were ranked and weighted in seven separate layers. In the next step, the subsidence location and rates were obtained using the differential interferometric synthetic aperture radar (DInSAR) method. The weights of the input parameters of the GARDLIF model using the subsidence map obtained from the DInSAR method and the particle optimization algorithm (PSO) were then optimized. Accordingly, the subsidence susceptibility map was generated based on the new weights.Findings & Conclusion The results showed that by increasing correlation coefficient (r) from 0.55 to 0.67 and the amounts of Coefficient of Determination (R2) from 0.39 to 0.53 between the subsidence index and the obtained subsidence in the aquifer, the optimization of weights applied by the PSO algorithm is more capable for evaluating the land subsidence than the map created by GARDLIF. It was also found that the central parts of the study aquifer had the largest potential for land subsidence.

A B S T R A C TA R T I C L E I N F O

Article TypeOriginal Research

AuthorsChatrsimab Z.1 MSc,Alesheikh A.*2 PhD,Vosoghi B.3 PhD,Behzadi S.4 PhD,Modiri M.5 PhD

Keywords DInSAR; GARDLIF; PSO; Subsidence; Vulnerability

*CorrespondenceAddress: No. 1346, ValiAsr Avenue.,Mirdamad Street, Tehran, Iran. Post-al Code: 1996715433Phone: +98 (21) 88786212Fax: +98 (21) 88786213alesheikh@kntu.ac.ir

1Department of GIS/RS, Science and Research Branch, Islamic Azad University, Tehran, Iran2GIS Engineering Department, K.N. Toosi University of Technology, Teh-ran, Iran3Geodesy Department, K.N. Toosi University of Technology, Tehran, Iran4Civil Engineering Department, Sha-hid Rajaee Teacher Training Univer-sity, Tehran, Iran5Geography Urban Panning Depart-ment, Malek Ashtar University of Technology, Tehran, Iran

Article HistoryReceived: August 10, 2019 Accepted: December 08, 2019 ePublished: May 19, 2020

How to cite this articleChatrsimab Z, Alesheikh A, Vosoghi B, Behzadi S, Modiri M. Land Subsi-dence Modelling Using Particle Swa-rm Optimization Algorithm and Diff-erential Interferometry Synthetic Aperture Radar. ECOPERSIA. 2020-;8(2):77-87.

LandSubsidenceModellingUsingParticleSwarmOptimization…78

ECOPERSIASpring2020,Volume8,Issue2

IntroductionGroundwater resources are one of the mostimportant sources of water supply [1] foragriculture, industry and, drinking in manyregionsofIran,especiallyinthecentral,easternand southern regions [2, 3], which led to over‐exploitation of groundwater from aquifers [4].Thus, themajor problems associated with theinappropriate extraction of undergroundaquiferswater are thepersistent groundwaterlevel declination and the sediments and layersaccumulation inaquifers [5],whichmay lead tolandsubsidence.Landsubsidence isoneoftheenvironmental problems and geologicalhazards that has been reported in more than150citiesof theworld [6].Thisphenomenon isderived from thenatural (including volcanoes,continentaldrift,acquisitionofsolutesthroughrainfall,andsoon)andman‐made(suchascoaland metal mining, and soil salts solutionthrough irrigation, over extraction ofgroundwater, oil or gas, and also excessconstruction) reasons [7, 8]. Generally, landsubsidencerateoccurredduetonaturalfactorsislessthanonecentimeterperyear,comparedwith theman‐made factorswhich reach 50cmperyear[9].The landsubsidence,whichhasbeenobservedinmanypartsof theworld,especially inmanyplainsofIran,ledtoendangerhumanlivesandresult in heavy financial burdens [10, 11]. Thiscauses topography deformation, damage tourban infrastructure and facilities, severeflooding, and a decrease in the capacity ofgroundwater aquifers to store water [12].Therefore, identification of effective factors,modelling, and mapping of the potential landsubsidence is very important for preventingsuch damages [13]. Kim et al. [14] using anartificial neural network and geospatialinformation system, predicted the landsubsidence in Samcheok of Korea. Theypresented thehazardmap for land subsidencewith 96% validation of field data andsubsidencelocationsinthearea.OhandLee[15]have been investigated land subsidence byGeographic Information System (GIS) and theWeights‐Of‐Evidence(WOE)modelusingsevenmainfactorsandreportedahighaccuracyratebetween the subsidence map and the formerlandsubsidencepositions.Theriskmappingofsubsidence using five parameters includingland slope, elevation, lithology, distance from

thevalley‐shaped region (sinkholes) and land‐use,hasobtainedbyPutraetal.[16].Theresultsindicated that the highest risk areas inaccordance with field information observednearthesinkholesregion.Inordertoprepareapotential subsidence map, Xu et al. [17]compared the Cosserat continuummodelwiththeCauchycontinuumclassicmodel.

Afifi [18] used Lamb‐Whitman experimentalrelation to evaluate the potential of landsubsidence and its effective parameters in theSeyedan‐Farooq Plain of Iran. The resultsshowed that the density and compressionamong clay layers and the inappropriateextraction of underground aquifers were themost effective factors occurring landsubsidence. Nadiri et al. [19] andManafiazar etal. [20] were estimated the potential of aquifersubsidenceusingthegeneticalgorithmandtheresult verified by subsidence obtained fromsatelliteimagery.Theresultsdemonstratedthatusing the genetic algorithm led to increase thecorrelation coefficient between the subsidenceindex and calculated subsidence in the plain.Manafiazar et al. [21] also used the ALPRIFTmethod and support‐vectormachine (SVM) toevaluate the subsidence vulnerability ofsouthwesternplain of Tehran. The coefficientsof the ALPRIFT model improved by the SVMmodel.Theresultsdepictedbetterefficiencyofthe SVM model for evaluating the subsidencevulnerability.

In the present study, a vulnerability map ofaquifersubsidencehasbeendevelopedusinganewGARDLIFmethod,providedbyNaderietal.[22]. In the next step, the particle swarmoptimization (PSO) algorithm was used tooptimize the coefficients and results of theGARDLIF method in Tehran‐Karaj‐ShahriyarAquifer. The simultaneous using satellite andterrestrial data is the other advantage of thisstudy, which provides the most available andreliable data. In overall, the current studyaimed to evaluate the efficiency of the PSOalgorithm in improving the coefficients ofparameters affecting subsidence susceptibilitymapaswellastodeterminethemostimportantparameters in occurrence of this phenomenonfor better landmanagement and its control inTehran‐Karaj‐ Shahriyar Aquifer. Regardingusing thePSOalgorithmtoprovideapotentialsubsidencemap, the present study is the firststudyinthisfield.

79ChatrsimabA.etal.

ECOPERSIASpring2020,Volume8,Issue2

MaterialsandMethodsCasestudy Tehran‐Karaj‐Shahriyar Aquifer with5083.97km2 width is located in a plain with2519.8km2 width. The study area is locatedbetween latitude 35°20´‐36°15´ N, andlongitude 50°50´‐52°15´ E. This aquifer wasprohibited by the Ministry of Energy of Iransince2008due to the inappropriateextractionof underground water. The maximum andminimum altitude in the study areawas 4375and 800m, respectively. The averagetemperature and rainfall at altitudes wererespectively11.4°Cand432.5mm/year,andtheaverage temperature and rainfall at the plainwere 16.2°C and 227.5mm/year, respectively(Figure1).

Figure 1) Location of Tehran‐Karaj‐Shahriyar Aquifer,Iran According to the last census of groundwaterresourcesin2011,40275wellswithanannualdischarge of 1931.74millionm3, 1336 springswith an annualdischargeof 108.45millionm3and 429 Qantas with an annual discharge of226.23 million m3, respectively, have beenreported. Water consumption in this areaincludes 2024.6 million m3 of groundwater(wells and Qantas) and 1013.9 million m3 ofsurface flowsandspring,which1394.17m3 foragricultural, 1587.41 million m3 for drinking

and 65872m3 for industry consumptions havebeenused.Growth in groundwater resource utilizationwasrelativelylowinthe1950sand1960s,andthen aquifer utilization rates increased fromyear1970onwards.Thehighestgrowthrateforaquifer exploitation related to 1999 to 2001years,with an average growth rate of about 5millionm3/year.Furthermore,accordingtothestatistics and information of piezometricwellsfrom October 1999 to 2015, the amount ofchanges in groundwater level was betweenzeroto‐40m.Consequently,thealluvialaquiferofTehran‐ShahriyarandKarajplainshasfallenby0.52monaverage.ResearchmethodologyThe modelling of potential subsidence ofTehran‐Karaj‐ShahriyarAquiferwasdoneusingpoint count system models, GARDLIF, particleswarm optimization algorithm (PSO) anddifferential interferometric synthetic apertureradar(DInSAR;Figure2).

Figure2)StepsofthestudyAt the first, the GIS technique has providedlayers of effective subsidence parametersincluding groundwaterdeclination (G), aquifermedia(A),netrecharge(R),discharge(D),landuse (L), impact of aquifer thickness (I), anddistanceoffault(F)fortheGARDLIFmodel.Regarding these methods, each effectivesubsidenceparameterhasweighed1‐5andthesub‐classesofeachparameterwereassignedaweightofbetween1‐10(Table1)accordingtotheirimportance(Equation1).(1)SPI=WGr+WAr+WRr+WPr+WLr+WTr+WFr

LandSubsidenceModellingUsingParticleSwarmOptimization…80

ECOPERSIASpring2020,Volume8,Issue2

Table1)WeightsofaffectingparametersandrankofeachsubclassintheGARDLIFmodel[19,20,21]

Aquifermedia(W=5)

Landuse(W=3) Pumping(W=4) Recharge(W=4)

Impactsofaquiferthickness(W=2)

Faultdistance(W=1)

Diclineofwatertable(W=5)

Subsidencevulnerabilityindices

Range Rate Range Rate Range(cm/y)Rate Range(cm/y)Rate Range Rate

Range(km) Rate

Range(m/y) Rate Class

Clay 10 Mine 9‐10 <0.0001 1 0‐4 10 0‐25 1 0‐1 10 0‐0.2 1 24‐78

Low

Clay+Silt 9 Agriculture 7‐9 0.0001‐0.005 2 4‐9 9 25‐55 2 1‐2 8 0.2‐0.5 2 78‐132

Moderate

Clay+Sand 8 Damsite 6‐9 0.005‐0.01 3 9‐14 7 55‐90 3 2‐3 6 0.5‐0.9 3 132‐186

High

Silt‐Clay+Sand

3‐5 Residential 3‐8 0.01‐0.5 4 14‐19 5 90‐130 4 3‐4 4 0.9‐1.4 4 186‐240

VeryHigh

Sand 4 Transportation 3‐4 0.5‐1 5 19‐24 3 130‐175 5 4‐5 2 1.4‐2 5 Gravel 3 Dryareas 1‐3 1‐5 6 >24 1 175‐225 6 >5 1 2‐2.7 6 Organicsoils 8‐10 Wasteland 1 5‐20 7 225‐280 7 2.7‐3.5 7

Rubble 2 Grassland 1 20‐40 8 280‐340 8 3.5‐4.4 8 40‐65 9 340‐405 9 4.4‐5.4 9 >65 10 >405 10 >5.4 10

Where SPI was a subsidence vulnerabilityindex, W’ denotes weights and followinguppercase acronyms represent the GARDLIFdatalayers,andsubscripts‘r’denotesrates. Sevenlayerswerecreatedbasedontheweightof the parameters and the weight of thesubclasses of each parameter in the GISenvironment.Then,byintegratingtheselayers,a subsidence sensitivitymapwas obtained bytheGARDLIFmethod. Next, the existing subsidence map of thisaquifer was then prepared using SAR radarimages from the ENVISAT sensor and theDInSARmethodbetweentimeintervals2003to2009. In order to optimize the coefficients ofthe GARDLIF method, the PSO algorithm wasused. The input to the PSO algorithmwas theweights given to the parameter affecting thesubsidence.The objective function was to maximize thePearson correlation coefficient between thesubsidence maps available in the study areaobtained by the DInSAR method and thesubsidence potential index obtained from theGARDLIFmodel.Finally,basedontheoptimumweights, a subsidence sensitivity map wasprepared and compared with the results ofDInSAR.ParametersofGARDLIFmodelAquifermediaFine‐grained soils (such as silt and clay) aremore compact compared with sand andpebbles. After underground water extraction,fine‐grained soils are going to undergo anirreversible consolidation and resulting tosubsidenceduetothelackofelasticityandhigh

consolidation coefficients [23]. Therefore,subsidenceoccursmore frequently in the areacontaining thick sedimentary deposits oraquiferlocatedbetweentheclayandsiltlayers[19]. The geological well logs related to thepiezometers in theplain,whichobtained fromthe Geological Survey andMineral ExplorationofIran(https://gsi.ir/fa),wereusedtopreparethis layer. Totally, 109well logs were used inthisstudyarea(Figure3).LanduseVarious landusesshoweddifferent impactsonsubsidencerate.Theweighingandrasterlayerswere prepared according to Table 1instructionsaswellasdepictedinFigure3.DischargeThere is a balance between recharge anddischarge of groundwater aquifer, butagricultural pumping and urban consumptionled to destroy this balance and resulting landsubsidence [24]. The most important cause ofland subsidence in the sedimentary basins isthe accumulation of groundwater aquifersdueto the excessive pumping of groundwaterresources [25]. The higher amount ofgroundwater extraction resulted in the lowerhydraulic pressure and decrease the spacebetween the seeds, which led to increase theeffective tension and increased the density ofthe layers, so the occurrence of subsidenceincreases[21].Thislayerwasprovidedusingtheannual extraction rate of wells located on theplain. The Thiessen polygon method wasappliedtodeterminetheexchangerateofeachpiezometer and then related layer wasprepared(Figure4).

81ChatrsimabA.etal.

ECOPERSIASpring2020,Volume8,Issue2

Figure3)Aquifermedia(1)andlanduse(2)mapsofthestudyarea

Figure4)Discharge(1)andrecharge(2)ratemapsofthestudyarea

(1)

(2)

(1)

(2)

LandSubsidenceModellingUsingParticleSwarmOptimization…82

ECOPERSIASpring2020,Volume8,Issue2

RechargerateThe recharge rate is the amount ofwater thatenters the aquifer from the surface [26]. Thehigher recharge rate led to increasing thehydraulic pressure and the distance betweenthe seeds, which results in decreasing theeffective tension and the occurrence ofsubsidence. The Piscopo method (2001) [27],which consists of combining of the slope, soilpermeability,andrainfalllayers,wasappliedtoprepare the layer of recharge rate, then theweighing and raster layers were preparedaccordingtoTable1(Figure4).AquiferthicknessAreaswithhighreductionoffreesurfacewaterand high thickness of the fine‐grained andaquifer layers showedgreater land subsidence[28]. This layerwaspreparedusing geo‐electricsectionsoftheplain,whichdatesbackto2007.After interpolatingandobtainingraster layers,therankingscalewasmadeaccordingtoTable1andthefinalthicknessmapoftheaquiferwasobtained(Figure5).DistanceoffaultGIS Euclidean distancewas used to obtain thedistance of fault for each point of the plain.Then the resultingmapwas ranked accordingtoTable1(Figure5).GroundwaterdeclinationIn order to provide this layer, the informationofpiezometersoftheregionduringsevenyearswas obtained from Tehran Regional WaterAuthority (http://wrbs.wrm.ir). The differencein water surface was obtained from October2001 to 2008. After interpolating the wholeregiondatausingtheKrigingmethod,therasterlayer was obtained to integrate with otherlayers. This layer was ranked according toTable 1 and eventually, the groundwaterdeclinationmapwasachieved(Figure6).Particle Swarm Optimization (PSO)algorithmThis algorithmwas proposed by Kennedy andEberhart in 1995 [29], who used an initialpopulation includingpotentialproblemsolvingto explore the search space. Themain idea ofthealgorithmisthattherewasthepossibilityofreaching the goal for each category and allmembers through their observations andexperiences [30]. The responsibility of changingtheparticlestodiscoveramongthesolutionsbythevelocityvectorofeachparticleisoneofthedifferences between this algorithm and thegeneticalgorithm[31].

Figure5)Aquiferthickness(1)anddistanceoffaultmapsofthestudyarea(2)

(1)

(2)

83ChatrsimabA.etal.

ECOPERSIASpring2020,Volume8,Issue2

Figure6)GroundwaterdeclinationmapofthestudyareaThe particle velocity in each step consists oftwo parts [32, 33]. The first part was the initialvelocity of a particle and the second part wasrelated to thepursuitofthebestexperienceofeachparticle andwith theotherparticles.Thecombinationofthesetwopartsleadstocreateabalanceinthesearches[34].TheEquation2wasusedtoupdatetheparticlevelocity.(2)V t 1 w V t C ∗ rand1 t ∗pbest t position t C ∗ rand2 t ∗gbest t position t

Whichinthisequation:pbest [t]: The best position of each particle attimetGbest [t]: The best position for each particleamongthewholeparticlesc1: Constant coefficient (the highest rate ofparticlemotiononthebestpath)c2: Constant coefficient of training (motion inthepathofthebestparticlefoundedinthetotalpopulation)rand1 and rand2: Two random numbers withuniformdistributionontheinterval0‐1V[t]:VelocityvectorattimetPosition[t]:PositionvectorattimetThe PSO algorithm was used to improve theresults and optimize the weights of theparameters affecting subsidence. The input

parameters consist of seven weightingparameters affecting the subsidence, whichwere entered in the model as the initialpopulation. The objective function was tomaximize the Pearson correlation coefficientbetweenthesubsidencesoccurring in theareaobtained using radar images and the DInSARmethod and the subsidence potentials indexderived from the integration of the layers(Equation3).(3)

r=∑ ̅

∑ ̅ ∑

Whichinthisequation:r: The objective function in the optimizationmodeln:Numberofmeasuredpointsxj:Vulnerabilityindexrelatedtopointjx:Averageofvulnerabilityindexyj:Subsidencerateatpointjy:Averageofsubsidenceratewi:WeightsappliedtoeachParameterThe condition for stopping this study wassimilar objective functions in severalrepetitions. Finally, the GIS technique hasprovidedtomaplayersusingthesecoefficientsand the final map of the subsidencevulnerabilitygeneratedusingthisalgorithm.Verification of results using DifferentialInterferometric Synthetic Aperture Radar(DInSAR)methodThe differential interferometric syntheticapertureradartechniquewasusedinthisstudyto examine the subsidence rate. At the firsttime, this method has been proposed byGoldstein and Zebker [35]. In the DInSARmethod, at least three inputs (twoSLCs of theareaandaDEMdigitalelevationmap)orthreeSLC images of the area demonstrated thedisplacementoccurson the land.Theaccuracyof thismethoddependedonthewavelengthofdata used which is equal to half of thewavelength(γ/2).In order to reduce the lack of correlation, thevalues of the spatial reference and timeline oftheimagesshouldbeconsidered.Thenextstepis geometric image recording from the sameposition,whichtheslaveimagerecordedonthemaster image. According to the phasedifference between the two radar images, thearea interferogram was developed using theslave and master image. Until this step, the

LandSubsidenceModellingUsingParticleSwarmOptimization…84

ECOPERSIASpring2020,Volume8,Issue2

geometric errorhasbeeneliminated, so in thenext step, the topographic error should beremoved and appropriate filters should beapplied.TheSRTM90‐mdigitalelevationmodel(DEM) was used in order to eliminatetopographic effects [36, 37]. At the final step, thephase correctionwas applied and themaps ofsubsidence rate and range were prepared(Figure 7). The ESAR Envisat satellite radardata at C bandwas used in the present study.The images processing were carried out bySARSCAPE5.3softwareontheENVIplatform.

Figure7)SubsidencemapobtainedfromDInSARmethodFindingsandDiscussionThe layers were combined based on theweightsassignedtoeach layer in theGARDLIFmodelandthevulnerabilitymapoftheTehran‐Karaj‐Shahriyar Aquifer was prepared (Figure8). The potential subsidence index wasestimatedintherangefrom74to188.ThePSOmodelwasappliedtooptimizeandevaluatetheweights used in the GARDLIFmodel, which isdefined according to the objective functionshowing the Pearson correlation between theGARDLIF subsidence potential and the radardata subsidence. The weight values of eachlayer were optimized (Figure 8). Table 2showed the optimized weights obtained fromthe PSO algorithm and the weights applied totheGARDLIFmodel.

Table2)TheweightofparametersaffectingsubsidenceModel GARDLIF PSOGroundwaterdeclination 5 5Aquifermedia 5 4.6Rechargerate 4 4.3Dischargerate 4 3.6Aquiferthickness 2 2.67Landuse 3 4.1Distanceoffault 1 0.2

Figure 8) Potential subsidence maps obtained fromGARDLIF(1)andPSO(2)methods

(1)

(2)

85ChatrsimabA.etal.

ECOPERSIASpring2020,Volume8,Issue2

EvaluatingcriteriatoassesstheaccuracyofthemodelsCorrelations index (CI) and the Pearsoncoefficient were used to compare the usedmodels. In order to use the CI, the subsidenceratewasdivided into fourcategories includingveryhigh,high,medium,andlow[19,20,38].Then,thenumbersofwells thatdemonstrated equalsubsidence rate in comparing with thesubsidencevulnerabilitymapsandalsolocatedin the same group were multiplied by 4. Thenumberofwellswhichshowedthedifferenceof3, 2, and 1 values subsidence rate than thesubsidencevulnerabilitymapsweremultipliedby 3, 2, and 1, respectively. Then, the sumvaluesandthecorrelationindexwereobtained.The result showed the good correlationbetween the occurrence of subsidence rate inthestudyareaandtheusedmodels.70piezometricwellswereused in thepresentstudy.After that, the subsidence rate obtainedby thesatellitemapwasalsodivided into fourcategories including very high, high, medium,and low. Table 3 showed the conformance ofpiezometersandvulnerabilitygroupsobtainedby the proposed framework and the PSOalgorithm. The conformance resultsdemonstrated that the PSO model showedbetter results with weight optimization thantheGARDLIFmodel.ThefinalresultspresentedinTable3andDiagram1.Based on the results, the highest subsidencerateaccordingtobothmethodsanddifferentialinterferometricmapof subsidenceoccurred inthe central part of the plain. Regarding theeffect of groundwater declination onsubsidence rate, the obtained results showedthat the highest groundwater declination hasbeen recorded in the central areaof theplain.Inlinewiththeobtainedresults,Razmgiretal.[39]alsostatedthattheamountofsubsidenceindifferent parts of Tehran‐Shahriyar Plain isdifferentandhasaVpatternwithamaximumsubsidence rate of about 16cm per 0.78m ofwater table fall in the central part. Over‐watering was reported as themost importantcauseofsubsidenceinthestudyarea.Saffarietal. [40] reported the maximum average annualsubsidenceof136mmintheperiod2003‐2010in the center of Shahriyar Plain wheremetropolitanarea,itsgardensandfarmlandarelocated.Themaincauseof thesubsidencewasrelated to the indiscriminate extraction ofgroundwater. In the plain area of the Tehran‐

Karaj‐Shahriyar Aquifer, the size of thesedimentswas reduced towards the center, sothatclaydepositsshowedasignificanteffectontheoccurrenceof subsidence in this sectionofplain. On the other hand, there are coarse‐grained sediments in the northern andsouthern parts of the plain which played animportant role in the recharge rate andwaterentranceintheaquifer.Awideareaofthestudyplain has been used for agricultural purposes,whichbasedontheinappropriateextractionofunderground aquifers water for irrigationshowed an important role in subsidence rate.The stated expressions have complied withareas of the high potential of subsidence(Figure8).Table3)ComparisonofaccuracycriteriaofGARDLIFandPSOmodelsSPIScoresVerylow Low Moderate CI r R2PSOModelVerylow 1

238 0.67 0.536Low 25 18 2Moderate 7 15GARDLIFModelLow 26 19 7 233 0.55 0.395Moderate 6 10

Diagram 1) Correlation coefficient (r) of SPI‐GARDLIF,SPI‐PSOandDInSARsubsidencerate

LandSubsidenceModellingUsingParticleSwarmOptimization…86

ECOPERSIASpring2020,Volume8,Issue2

ConclusionTehran‐Karaj‐Shahriyar Aquifer is located inTehran, Shahriyar,Karaj andFashafuyeplains.Increasingpopulation and demand for variouswater uses led to increase extraction ofunderground aquifers water during the pasttwodecades,whichresultedinasharpdeclinein water surface and aquifer volume.Undoubtedly, the continuations of theseconditionswillcauseirreparableconsequences.Therefore, the study and assessment of thepotentialofsubsidenceareessentialinordertomanage the destructive effects of landsubsidence.Towardsthis, inthepresentstudy,seven parameters affecting subsidence wereused in theGARDLIFmodel.Then, theparticleswarm algorithm (PSO) was used to optimizethe model coefficients. The results of thepresent study showed that themethod of PSOalgorithm due to increase the correlationbetween normalized subsidence rates andnormalized subsidence potentials,demonstrated the vulnerability subsidence ofthe area better than the GARDLIF method.Basedonthismodel,thenorthernandsouthernparts of the plain showed a greater risk ofsubsidence, so management programs shouldbedonetocontrolandprotecttheseareas.Acknowledgments: This study was supported byIslamic Azad University, Tehran Science andResearch Branch, Islamic Republic of Iran. Theauthors are grateful to En. Mohammad HosseinGhavimipanah (Sari Agricultural Sciences andNatural Resources University) and Dr. ZeinabHazbavi (University of Mohaghegh Ardabili), fortheir assistance in improving the text andanonymous reviewers for their useful commentsandreviews.EthicalPermissions:Nonedeclaredbytheauthors.ConflictofInterests:Theauthorsstatethatthereisnoconflictofinterest.Authors’ Contribution: Chatrsimab Z. (Firstauthor), Introductionauthor/Methodologist/Original researcher/Dataanalyst(30%);AlesheikhAA.(Secondauthor),Dataanalyst (25%). Vosoghi B. (Third author),Introduction author/Methodologist/Data analyst(20%); Behzadi S. (Fourth author), Introductionauthor/Data analyst/Discussion author (15%);Modiri M. (Fifth author), Methodologist/Assistantresearcher(10%)Funding/Support: The present study wassupported by Islamic Azad University, TehranScience and Research Branch, Islamic Republic ofIran.

References1‐ Motiee H, McBean E. Assessment of climate changeimpactsongroundwaterrechargefordifferentsoiltypes‐guelph region in grand river Basin, Canada. Ecopersia.2017;5(2):1731‐44.2‐AlijaniR,VafakhahM,MalekianA.Spatialandtemporalanalysis of monthly stream flow deficit intensity inGorganroud watershed, Iran. Ecopersia. 2016;4(1):1313‐29.3‐MotaghM,WalterTR,SharifiMA,FieldingE,SchenkA,Anderssohn J, et al. Land subsidence in Iran caused bywidespread water reservoir overexploitation. GeophysResLett.2008;35(16):L16403.4‐ Bazrafshan O, Zamani H, Etedali HR, Dehghanpir S.Assessment of citrus water footprint components andimpact of climatic and non‐climatic factors on them. SciHortic.2019;250:344‐51.5‐BazrafshanO,ParandinF, FarokhzadehB.Assessmentof hydro‐meteorological drought effects on groundwaterresources inHormozgan region‐South of Iran.Ecopersia.2016;4(4):1569‐84.6‐HuR,WangS,LeeC,LiM.Characteristicsandtrendsofland subsidence in Tanggu, Tianjin, China. Bull Eng GeolEnviron.2002;61(3):213‐25.7‐BouwerH.Groundwaterhydrology. NewYork:McGrawHill College;1978.8‐. MalekiA,RezaeiP.ForecastlocationsatriskofsubsidenceplainKermanshah.ModarresHumSci.2016;20(1):235‐51.[Persian]9‐KuehnF,AlbiolD,CooksleyG,DuroJ,GrandaJ,HaasS,etal.DetectionoflandsubsidenceinSemarang,Indonesia,using stable points network (SPN) technique. EnvironEarthSci.2010;60(5):909‐21.10‐TaheriTizroA,HosseiniA,KamaliM.Modelingalluvialaquifer using PMWIN software and evaluation ofsubsidence phenomenon in Asadabad plain, HamedanProvince, Iran. Nat Environ Hazards. 2018;7(17):121‐36.[Persian]11‐ Hazbavi Z, Sadeghi SH. Watershed healthcharacterization using reliability‐resilience‐vulnerabilityconceptual framework based on hydrological responses.LandDegradDev.2017;28(5):1528‐37.12‐ Bhattarai R, Alifu H, Maitiniyazi A, Kondoh A.DetectionoflandsubsidenceinKathmanduValley,Nepal,usingDInSARtechnique.Land.2017;6(2):39.13‐FultonA.Landsubsidence:Whatisitandwhyisitanimportant aspects of groundwater management?.Sacramento: California Department of Water Resources;2006.14‐KimK,LeeS,OhHJ.Predictionofgroundsubsidencein Samcheok City, Korea using artificial neural networksandGIS.EnvironGeol.2009;58(1):61‐70.15‐OhHJ,LeeS.AssessmentofgroundsubsidenceusingGIS and the weights‐of‐evidence model. Eng Geol.2010;115(1‐2):36‐48.16‐PutraDP,SetiantoA,KeokhampuiK,FukuokaH.Landsubsidenceriskassessmentcasestudy:Rongkop,GunungKidul, Yogyakarta‐Indonesia. The 4th AUN/SEED‐NetRegionalConferenceonGeo‐DisasterMitigation inASEA,2011 October 25‐26, Thailand. Unknown Publisher city;UnknownPublisher;2011.17‐ Xu YS, Yuan Y, Shen SL, Yin ZY, Wu HN,  Ma L.InvestigationintosubsidencehazardsduetogroundwaterpumpingfromAquiferIIinChangzhou,China.NatHazard.2015;78(1):281‐96.

87ChatrsimabA.etal.

ECOPERSIASpring2020,Volume8,Issue2

18‐AfifiMA.Assess thepotentialof land subsidenceandits related factors (Case study: Plain Saidan FaroukMARVDASHT).QuantGeomorpholRes.2017;5(3):121‐32.[Persian]19‐Nadiri AA, Taheri Z, Khatibi R, Barzegari G, DidebanKh.Introducinganewframeworkformappingsubsidencevulnerability indices (SVIs): ALPRIFT. Sci Total Environ.2018;628:1043‐57.20‐ManafiazarA,KhamehchiyanM,NadiriA.ComparisonofvulnerabilityofthesouthwestTehranplainaquiferwithsimple weighting model (ALPRIFT Model) and geneticalgorithm(GA).KharazmiJEarthSci.2019;4(2):199‐212.[Persian]21‐Manafiazar A, KhamechianM,NadiriA. Optimizationof the ALPRIFTmethod using a support vector machine(SVM) to assess the subsidence Vulnerability of thesouthwestern plain of Tehran. J Eng Geol. 2018;11(2):1‐14.[Persian]22‐NaderiK,NadiriAA,AsghariMoghaddamA,KordM.Anew approach to determine probable land subsidenceareas (Case study: The Salmas plain aquifer). Iran JEcohydrol.2018;5(1):85‐97.[Persian]23‐ Bouwer H. Groundwater hydrology. Lotfi‐Sadigh A,translator.Tabriz:SahandUniversityofTechnologyPress;1995.[Persian].24‐ Galloway DL, Burbey TJ. Review: Regional landsubsidence accompanying groundwater extraction.HydrogeolJ.2011;19(8):1459‐86.25‐PachecoJ,ArzateJ,RojasE,ArroyoM,YutsisV,OchoaG. Delimitation of ground failure zones due to landsubsidenceusinggravitydataandfiniteelementmodelingin the Queretaro valley, Mexico. Eng Geol. 2006;84(3‐4):143‐60.26‐ Alizadeh A. Principles of applied hydrology. 9thEdition. Mashhad: Imam Reza university Press; 1996.[Persian].27‐ Piscopo G. Groundwater vulnerability mapexplanatorynotes.Parramatta:NSWDepartmentofLandandWaterConservation;2001.28‐HafezimoghadasN,GhafooriM.EnviromentalGeology.1stEdition.Shahrood:ShahroodUniversityofTechnologyPress;2009.[Persian].29‐KennedyJ,EberhartRC.Particleswarmoptimization.Proceedings of ICNN'95‐International Conference onNeuralNetworks,199527November‐1December,Perth,Australia.Piscataway:IEEE;1995.30‐EnshaeeA,HooshmandR.Applicationoffuzzyparticleswarm optimization in detection and classification of

singleandcombinedpowerqualitydisturbances. ModaresJElectricEng.2010;10(2):1‐16.[Persian]31‐MoghaddasiM,MoridS,AraghinejadSh.Optimizationofwaterallocationduringwater scarcity conditionusingnon‐linear programming, genetic algorithm and particleswarm optimization (case study). J Water Res Agric.2009;4(3):1‐13.[Persian]32‐ Taherifar A, Alasty A, Salarieh H, Boroushaki M.Path planning for a planar hyper‐redundantmanipulator with lockable joints using particle swarmoptimization.  Modares Mech Eng. 2011;11(2):159‐75.[Persian]33‐ Shafiei Alavijeh M, Amirabadi H. Modeling andoptimizing lapping process of 440C steel by neuralnetworkandmulti‐objectiveparticleswarmoptimizationalgorithm. Modares Mech Eng. 2017;17(8):201‐12.[Persian]34‐AlizamirM,AzhdaryMoghadamM,HashemiMonfaredA, Shamsipour A. A hybrid artificial neural network andparticle swarm optimization algorithm for statisticaldownscaling of precipitation in arid region. Ecopersia.2017;5(4):1991‐2006.35‐ZebkerHA,GoldsteinRM.Topographicmappingfrominterferometric synthetic aperture radar observations. JGeophysRes.1986;91(B5):4993‐9.36‐ AhmadW, Choi M, Kim S, Kim D. Detection of landsubsidence due to excessive groundwater use varyingwithdifferent landcovertypes inQuettavalley,Pakistanusing ESA‐sentinel satellite data. NatHazards Earth SystSciDiscusss.2017;1‐21.37‐CalóF,NottiD,GalveJP,AbdikanS,GörümT,PepeA,et al. Dinsar‐Based detection of land subsidence andcorrelation with groundwater depletion in Konya Plain,Turkey.RemoteSens.2017;9(1):83.38‐RahmatiO,FalahF,NaghibiSA,BiggsT,SoltaniM,DeoRC, et al. Land subsidence modelling using tree‐basedmachine learning algorithms. Sci Total Environ.2019;672:239‐52.39‐ Razmgir R, Mousavi M, Shemshaki A, Bolourchi MJ.Tehran‐ShahriarPlainsubsidenceduetoexcessextractionof underground water, preliminary survey. 1st NationalConference on Coastal Water Resources Management,2010 December 8, Sari, Iran. Sari: Sari University ofAgricultural Sciences and Natural Resources; 2010.[Persian]40‐ Safari A, Jafari F, Tavakoli Sabour S. Monitoring itsland subsidence and its relation to groundwaterharvesting case study: Karaj plain‐Shahriar. QuantGeomorpholRes.2016;5(2):82‐93.[Persian]