Assessments of Serious Anthropogenic Land Subsidence in Yunlin County of Central Taiwan from 1996 to 1999 by Persistent Scatterers Insar

Tectonophysics 578 (2012) 126–135

Contents lists available at SciVerse ScienceDirect

Tectonophysics

journal homepage: www.elsevier.com/locate/tecto

Assessments of serious anthropogenic land subsidence in Yunlin County of central Taiwan from 1996 to 1999 by Persistent Scatterers InSAR

Hsin Tung, Jyr-Ching Hu ⁎

Department of Geosciences, National Taiwan University, Taipei, Taiwan article info abstract

Article history: Anthropogenic ground subsidence due to massive pumping of groundwater is one of the severe environmen- Received 30 November 2010 tal hazards in Taiwan. The Yunlin County located in the southwestern coastal region of Taiwan is one of the Received in revised form 9 August 2012 most counties with serious land subsidence because of the agricultural needs. In certain areas of the region, Accepted 9 August 2012 the subsidence rate reaches as much as 14.3 cm/yr. The severe land subsidence gives rise to the risk of ﬂood Available online 18 August 2012 hazard and damage of infrastructures in this area. We represented a Persistent Scatterers InSAR (PSInSAR) results deduced from 1996 to 1999 time span for monitoring of land subsidence in this area. The PSInSAR Keywords: Land subsidence results show that Baojhou, Tuku and Yuanchang Townships reveal a maximum subsidence rate of about PSInSAR 7.8 cm/yr along the LOS and Lunbei Township located on the northern Yunlin reveals a subsidence rate of Precise leveling 3.5 cm/yr, which is quite coincident with the precise leveling result. This result has proven that the effective Groundwater pumping reduction of labor and cost could be achieved by using this technique on monitoring land subsidence in Natural hazards Yunlin County. © 2012 Elsevier B.V. All rights reserved.

1. Introduction underground water has resulted in environmental hazard and potential risk in Taiwan which the severe land subsidence of about 1 to 10 cm/yr The anthropogenic activities of excessive groundwater utilization was observed in several counties from 2002 to 2006 (Fig. 1b, data from in agriculture, aquaculture, industry and urban area give rise to Water Resource Agency, Department of Economics). Particularly in the serious land subsidence. Consequently, land subsidence could result in Choshui River alluvial fan where the Yunlin section of the Taiwan High environmental hazards such as exhaustion of groundwater resources, Speed Rail (THSR) had been constructed through the central of subsi- damage of infrastructures, increase of risks of inundation and inland dence area which might pose a serious threat of its operation (Chang sea water intrusion (Abidin et al., 2001; Amelung et al., 1999; Chen and Wang, 2006; Hwang et al., 2008). From 1992 to 2007, a severe et al., 2007; Hou et al., 2005; Hsieh et al., 2011; Hung et al., 2011; cumulative land subsidence larger than 110 cm was observed at Motagh et al., 2007; Osmanoglu et al., 2010; Phien-wej et al., 2006; Mailiao, Lunbei and Baojhon Townships (Fig. 2). From 1996 to 1998, Teatini et al., 2005; Wang et al., 2011). In Taiwan, groundwater has the center of land subsidence was located in Lunbei and Baojhon been abundantly used as an alternative to surface water, especially in Townships with a subsidence rate of 7–8 cm/yr (Fig. 3a). However the the southwestern coastal region where the deﬁciency of surface water center of land subsidence changed to Tuku and Yuanchang counties resources is severe due to the high water demand from aquacultural and almost covered the THSR (Fig. 3b). It is believed that the land sub- and industrial utilization in Taiwan (Hsu, 1998). From a tectonic view- sidence is caused by a deformation of clay or sand layers by compres- point, Taiwan is situated along the ongoing collision boundary sion, accompanied by heavy withdrawal of underground water near between the Eurasian Plate and the Philippine Sea Plate (Fig. 1a) with the coastal regions of Yunlin County (Liu et al., 2004). A lot of efforts a convergence rate of about 8 cm/yr (Hu et al., 2001; Lin et al., 2010). have been done in managing groundwater over-extraction and land Most of the western coastal plains are considered to be in the foreland subsidence in the coastal area (Hsu, 1998; Tang and Tang, 2006). Sea- basin of the Taiwan orogen (Lin and Watts, 2002). Although the mor- sonal effects of land subsidence occurring in the study area had been es- phology of western Taiwan is ﬂat, the subsurface deformation is charac- timated using a regression analysis of a series of weekly GPS height terized by various subsiding or uplift rate in response to the incipient solutions. The average rate of ground subsidence in this area over the development of fold-and-thrust belt (Huang et al., 2006a). However, period of 1995–2001 was 3 cm/yr (Chang and Wang, 2006). Based on anthropogenic ground subsidence induced by heavy withdrawal of the data collected at the piezometer, the variation of land subsidence rate appeared to be associated with an unstable underground water level, which drops gradually during winter and either remains constant ⁎ Corresponding author. Tel.: +886 2 23634860; fax: +886 2 23636095. or rises during summer time. Consequently, land subsidence rates vary E-mail address: [email protected] (J.-C. Hu). considerably from 1.5 cm/yr for the summer time to 9.0 cm/yr for the

Fig. 1. (a) Topography, bathymetry and main geological units in Taiwan. Rectangle indicates the study area. The large red indicator arrow notes direction and convergence rate of Philippine Sea Plate relative to Eurasia Plate. Major thrust faults with triangles are on the upthrust side. (b) Average land subsidence area in Taiwan during 2002 to 2006. Black lines are the administrative boundaries. Data from Water Resource Agency, Department of Economics. 127 128 H. Tung, J.-C. Hu / Tectonophysics 578 (2012) 126–135

Fig. 2. Cumulative land subsidence in Yunlin County by precise leveling from 1992 to 2007. Maximum cumulative land subsidence >110 cm occurred in Baojhon, Tuku and Yuanchang Townships. Red line indicates Taiwan High Speed Rail (TSHR). Data from Water Resource Agency, Department of Economics.

winter time (Chang and Wang, 2006). In addition, the annual subsi- from SAR interferograms (e.g., Buckley et al., 2003; Ding et al., 2004, dence rate is deeply influenced by annual rainfall. An annual subsidence 2008). Thus persistent scatterer (PS) technique has been proposed to rate of greater than 3 cm/yr is considered significant in the Choshoui improve the applicability of radar interferometry when applied to River alluvial fan. Based on the six leveling surveys carried out in detect long-term ground deformation with tracking the signals of dis- 2000, 2002, 2003, 2005, 2006, and 2007, the maximum annual subsi- crete point-wise targets (Berardino et al., 2002; Ferretti et al., 2000, dence rate was 9.5 cm/yr in 2002, 12.2 cm/yr in 2003, 11.6 cm/yr in 2001; Hooper et al., 2004; Kampes and Hanssen, 2004; Liu et al., 2005, 10.1 cm/yr in 2006 and 8.2 cm/yr in 2007, respectively (Hung 2008; Mora et al., 2003). In this paper, we use PSInSAR technique to et al., 2010). The subsidence rate was highest in 2003 due to a serious deduce the perturbation and obtain land subsidence motion around a drought that occurred in Taiwan which led to excessive pumping of section of THSR in Yunlin County in central Taiwan by the natural tar- groundwater. According to the report of Hung and Liu (2007),the gets recognized from a time series of SAR interferograms. Thus 33 total area with a subsidence rate >3 cm/yr was 1600 km2 in 2001. ERS-1 and ERS-2 images (Table 1) acquired from 1996 to 1999 are The total area with a subsidence rate >3 cm/yr was reduced to employed to improve the monitoring density and to further character- 803 km2 in 2007 after the efforts of mitigations in land subsidence ize the land subsidence in the area of interest. since 2007. However, the Changhua and Yunlin areas still suffer with the problems of heavy land subsidence. Recent study using multiple 2. Geological setting and data sensor of multi-level compaction monitoring well demonstrates that the aquifer-system compaction occurs mostly below depths >200 m 2.1. Geological and hydrological background and could occur at the depth greater than 300 m (Hung et al., 2010). Although the observations from GPS measurements, precise leveling The study area is located in central Taiwan island (Fig. 1), which is a and monitoring wells provided robust monitoring of land subsidence zone of active continental deformation located at the plate boundary hazard, however spatial coverage is quite sparse due to these point zone of the Eurasian Plate (EUP) and the Philippine Sea Plate (PSP) observations, thus the low spatial density over large areas could be (Lin et al., 2010). The PSP moves towards the northwest with respect improved by the detection of surface deformation revealed by differen- to the stable EUP (Penghu Islands) at a rate of 82 mm/yr and the polar- tial SAR interferometry (D-InSAR). ity of subduction between the EUP and PSP is flipped near the central The InSAR technique has proven to be capable of measuring topo- part of Taiwan (Lin et al., 2010). Yunlin County is one of the important graphic and crustal deformation at fine space resolution of tens of agricultural production regions located in the southwestern coastal re- meters over wide coverage (e.g., Buckley et al., 2003; Bürgmann et al., gion of Taiwan where the irrigated area is up to 123,000 ha and the 2000; Ding et al., 2004; Huang et al., 2006b, 2009; Massonnet and agricultural water consumption reaches approximately 90% of all avail- Feigl, 1998; Pritchard and Simons, 2002; Wright, 2002; Yen et al., able water resources in the Choshui River Basin (Zhang, 2005). Due to 2008). However, temporal and spatial decorrelations of radar signal there is no sufficient surface water supply, the groundwater becomes have prevented this technique from more frequent utilization. another necessary source for water consumption. Besides, the accuracy of InSAR measurements may also be significantly Choshui River is the longest river in Taiwan. Choshui River alluvial reduced by atmospheric phase artifacts that are difficult to be removed fan covers a total area of 2000 km2 and bounded by Wu River in H. Tung, J.-C. Hu / Tectonophysics 578 (2012) 126–135 129

Fig. 3. Land subsidence in Yunlin County revealed by precise leveling from 1996 to 1999. (a) Maximum land subsidence of 7–8 cm/yr from 1996 to 1998 occurred in Mailiao, Lunbei and Baojhon Townships. Red line indicates Taiwan High Speed Rail (TSHR). (b) The maximum land subsidence areas with a rate of 7–8 cm/yr shift to Tuku and Yuanchang Townships from 1998 to 1999. Data from Water Resource Agency, Department of Economics. north, Pekang River in south, western Foothills in east and Taiwan and mudstone (Fig. 4a). The thickness of the sediments is about 750– Strait in west (Fig. 4a). The sediments in Choshui River alluvial fan 3000 m, and the mean grain size shows a tendency to decrease from originate from the rock formation in western Foothills and Central east to west. Thus, near the head the alluvial fan mainly compose of Range, which includes metamorphic quartzite, slate, shale, sandstone gravel and coarse sand, meanwhile the soil and ﬁne sand are mainly 130 H. Tung, J.-C. Hu / Tectonophysics 578 (2012) 126–135

Table 1 used to form 32 interferometric pairs (Table 1). The interferometric Descending orbit data processed for the Yunlin area (track: 232, frame: 3141). Baseline processing is carried out with the Diapason software developed by is relative to 19960131 (orbit: 23767). the Centre National d'Etudes Spatiales (CNES). The external DEM

Orbit Date Sensor Baseline (m) FDC (Hz) used to remove the topographic component of the interferometric 23767 1996/01/31 ERS-1 923 32.11 phase is the free distribution DEM from NASA's SRTM mission. In par- 4094 1996/02/01 ERS-2 1068 −4.40 ticular, the Delft Institute for Earth-Oriented Space Research (DEOS) 24268 1996/03/06 ERS-1 1171 30.87 precise orbital data was injected in the interferometric processing to − 5097 1996/04/11 ERS-2 967 8.71 further remove the orbital uncertainties. 4595 1996/03/07 ERS-2 1133 −15.69 24770 1996/04/10 ERS-1 1026 35.24 25271 1996/05/15 ERS-1 1075 33.31 3. Persistent scatterer analysis 6099 1996/06/20 ERS-2 425 −8.32 5597 1996/05/16 ERS-2 958 −8.22 The Persistent Scatterers (PS) InSAR is an advanced technique in 7100 1996/08/29 ERS-2 95 −17.07 comparison with conventional InSAR technique, which addresses to 8604 1996/12/12 ERS-2 381 −9.42 9606 1997/02/20 ERS-2 547 −3.94 overcome the problems of decorrelation for generating a time series 8104 1997/01/16 ERS-2 741 −13.20 of phase changes without atmospheric and DEM residual effects by 10106 1997/03/27 ERS-2 812 −3.96 computing only on sparsely distributed PSs which are pixels coherent − 10607 1997/05/01 ERS-2 350 10.16 over long time series. This technique has been developed in the late 11108 1997/06/05 ERS-2 490 −7.98 11609 1997/07/10 ERS-2 525 −6.78 1990s by A. Ferretti, F. Rocca, and C. Prati of the Technical University 12111 1997/08/14 ERS-2 782 −14.09 of Milan (POLIMI). The first algorithm to find out the PS pixels was 13613 1997/11/27 ERS-2 900 −16.47 brought up by Ferretti et al. (2000, 2001), and trademarked it as 14114 1998/01/01 ERS-2 967 −14.54 the “Permanent Scatterer technique™”. After that, similar processing − 14616 1998/02/05 ERS-2 453 28.00 algorithms have since been developed by Crosetto et al. (2003) and 15117 1998/03/12 ERS-2 408 −12.14 16118 1998/05/21 ERS-2 719 4.81 Kampes (2005). Besides, the SBAS (Small Baseline subset) technique 16620 1998/06/25 ERS-2 −335 −0.048 developed by Berardino et al. (2002) and StaMPS (Stanford Method 17121 1998/07/30 ERS-2 834 −0.22 for Persistent Scatterers) developed by Hooper et al. (2004) have 17622 1998/09/03 ERS-2 981 1.58 the same idea with that of PS-InSAR technique but with different 18623 1998/11/12 ERS-2 1040 −4.97 19125 1998/12/17 ERS-2 233 −10.89 names. 19625 1999/01/21 ERS-2 1028 −7.32 Persistent scatterers technique uses the largest contributor signal 20627 1999/04/01 ERS-2 309 −6.54 (i.e. bridges, buildings) as the signal of a resolution element which 20126 1999/02/25 ERS-2 1366 −17.26 we call persistent scatterers. In the time series, these PS points are 21128 1999/05/06 ERS-2 1122 0.12 stable enough that could refer the information in the whole area. 21630 1999/06/10 ERS-2 715 −0.78 The steps of the persistent scatterers technique can be processed as follows: 1) the appropriate master image selection, 2) differential interferograms formation, 3) stable point target selection, and 4) linear in the toe of the alluvial fan. Due to frequent flooding and channel deformation and DEM residual extraction. migration along the Choshui River, the complex inter-bedded or lens-structural clay, fine sand, medium fine sand, coarse sand and 3.1. Master image choosing and PS identification gravel layers are found in the unconsolidated formation in the flood plan (Fig. 4b). Liu et al. (2001, 2004), proposed that the Choshui The first step is to choose one SAR image that minimizes the sum River alluvial fan can be divided into four marine sequences (aquitard decorrelation, i.e., maximizes the sum correlation, of all the interfero- I–aquitard IV) and four non-marine sequences (aquifer I–aquifer IV) grams in the whole data set. The correlation depends on four terms: at a depth of 0 to 300 m (Fig. 4c). temporal interval (T), perpendicular spatial baseline (B⊥), Doppler centroid frequency baseline (FDC) and thermal noise (Hooper, 2006; Zebker and Villaseno, 1992): 2.2. Precise leveling measurements

ρ ¼ ρ ρ ρ ρ In order to monitor the subsidence rate, the government has total temporalspatialdopplerthermal ð Þ carried on the leveling measurements in this area for more than two ≈ − T − B⊥ − FDC ρ 1 1 f c 1 f c 1 f c temporal decades. According to the precise leveling measurement maintained T B⊥ FDC by the Industrial Technology Research Institute, the serious subsidence rate is well known to occur along the southwestern coast of Yunlin where ρ denotes correlation and superscript c denotes the critical County, and the average subsidence rate is about 10 cm/yr before value which means the limit of producing a useable interferogram. 1996. The most subsidence area transfer happened in inside Yunlin c c For ERS data, T=5 years, B⊥ =1100 m, F =1380 Hz, and ρ is County with a maximum land subsidence of 7–8 cm/yr during the DC thermal a constant value. period of 1996 to 1999 (Fig. 3). At the same time, the maximum cumu- In the processes of PSInSAR, the algorithm is established based on lative subsidence along the coastal area has decreased from 5 to 7 cm/yr the PS points selected. Thus, the selections of PS points are quite to 3–5 cm/yr observed in Taishi and Sihhu townships. In addition, the important. We used spatial coherence method to pick the point precise leveling data measured after 1999 indicate that the maximum targets. In order to assure the phase is stable enough, the pixel ex- subsidence center is located inside Yuanchang and Tuku Townships in hibits coherence always greater than the threshold in all data set Yunlin County with a rate of 7–8cm/yr. will be discriminated. All these pixels with this character will be selected as PS candidates. After picking up the PS candidates, we only 2.3. SAR datasets analyze these pixels in the following steps. It is important to point out that the PS candidates we selected by the above-mentioned algo- In this case, 33 C-band (radar wavelength of 5.6 cm) radar images rithms were the primary result, not the ﬁnal PS points. In the follow- collected by the ERS-1 and ERS-2 satellites of the ESA along descending steps, we will reduce the points with bad correction to the real ing orbits from track 232 between January 1996 and June 1999 were deformation to get the real PS points. H. Tung, J.-C. Hu / Tectonophysics 578 (2012) 126–135 131

Fig. 4. (a) Regional geological map of the Choshui River alluvial fan superimposed with 40 m digital elevation model. H: Recent; Q: Pleistocene to Pliocene; PM: Pliocene to Miocene; OE: Oligocene to Eocene; MI: Miocene. (b) Two hydrogeological profiles of the Choshui River alluvial fan. Profile lines are indicated in Fig. 4a. (c) Conceptual hydrogeological profile of the Choshui River alluvial fan at the profile b–b′. (a) Modified from CGS (1999), (b) modified from Liu et al. (2004), and (c) modified from Liu et al. (2004).

The mean coherence value of the 33 coherence values is calculated 3.2. Extraction of linear deformation for each pixel. We consider a pixel as a PS candidate if the mean coherence value of the pixel is larger than the given threshold of When producing a conventional DInSAR result by two SAR images, 0.3, and thus resulting in 5605 PS points for the study area (Fig. 5). the phase can be written as the sum of four terms: After selection of all the PSs, we establish a triangular irregular network (TIN) as shown in Fig. 5 from the identiﬁed PSs by the Delaunay ϕ ¼ ϕ þ ϕ þ ϕ þ ϕ ð Þ triangulation method (Liu et al., 2009). diff mov topo error atm noise 2 132 H. Tung, J.-C. Hu / Tectonophysics 578 (2012) 126–135

Fig. 5. The distribution of 5605 PSs and generated Delaunay triangular irregular network. All PSs are marked by red solid circles and triangular irregular network marked by green lines. where where

ϕmov phase change due to movement of the pixel in the range x,y position coordinates of the pixel; distance; Ti temporal baseline of the ith interferogram; ϕtopo error residual topographic phase due to misﬁt with the DEM; β nonlinear component of velocity; ϕatm phase due to atmospheric artifacts; a atmospheric artifacts; ϕnoise noise term due to temporal and spatial decorrelation, ther- n noise. mal noise and coregistration errors. Using the information we already know in Eqs. (2) and (3),a

Here, the movement term (ϕmov) has two contributions: linear model with linear movement and DEM error can be assumed as: movement and nonlinear movement, the linear movement only cor- π relates with velocity, and the DEM error is proportional to the per- δϕ ðÞ¼; ; ; ; 4 ½ðÞ; − ðÞ; model xa ya xb yb Ti λ ·Ti· vmodel xa ya vmodel xb yb pendicular spatial baseline: π ðÞ : þ 4 BTi ½ε ðÞ; −ε ðÞ; λ · ðÞ ðÞθ · model xa ya model xb yb π rTi · sin i ϕ ¼ ϕ þ ϕ ¼ 4 þ ϕ ð Þ mov linear nonlinear λ ·v·T nonlinear 3 ð7Þ 4π B ·ε ϕ ¼ ⊥ : ð Þ For the 32 interferograms in this study, we can form 32 observa- topo error λ · θ 4 r· sin tion equations for each connection based on Eq. (7), where there are two unknowns (i.e., Δh and Δv) to be estimated. Then, use the Because the phase of individual pixel has an inﬁnite solution (the following model coherence function to perform maximization γ phase is wrapped), the Delauney triangulation is then being used to (Ferretti et al., 2000): connect all the PSs. This kind of triangulations relates all the neigh- boring pixels of irregularly gridded data generating nonoverlapped hi 1 XN triangles. Differential operation of the interferometric phase values γ ¼ exp j· δϕ ðÞT −δϕ ðÞT ð8Þ model N diff i model i along each side (referred to as connection hereafter for simplicity) i¼0 of the triangles is carried out to reduce the effects of spatially correlat- ed errors such as the residual atmospheric effects and orbital errors. where N is the number of interferograms. This function is equal to 1 when the adjustment to the data is perfect, and zero in case of total Thus, the phase difference between two PSs (xa,ya) and (xb,yb)ona connection can be expressed as: decorrelation. Once the maximization process has been done for each connection, all the connections with coherence below a threshold are rejected. A least mean square process is necessary to obtain δϕ ¼ δϕ þ δϕ þ δϕ þ δϕ : ð Þ diff ε mov atm noise 5 absolute values for each point.

Δv ðÞ¼x ; y ; x ; y ½v ðÞx ; y −v ðÞx ; y For the whole data set, it becomes as: estimated a a b b model a a model b b maximize γ ð Þ Δε ðÞ¼; ; ; ½ε ðÞ; −ε ðÞ; 9 estimated xa ya xb yb model xa ya model xb yb maximize γ 4π δϕ ðÞ¼; ; ; ; ½ðÞ; − ðÞ; γ γ∈ diff xa ya xb yb Ti λ ·Ti· vxa ya vxb yb denotes the overall coherence (OC) of a connection ( p[0,1]);ffiffiffiffiffiffiffiffi M is π ðÞ the number of interferometric pairs (33 in this case); j ¼ −1; and ε þ 4 BTi ½εðÞ; −εðÞ; k λ · ðÞ ðÞθ · xa ya xb yb ð Þ Δε Δ rTi · sin i 6 is the residual. In practice, both and v are determined by þ½βðÞ; ; −βðÞ; ; xa ya Ti xb yb Ti searching a predefined solution space to maximize the OC value. It þ½ðÞ; ; − ðÞ; ; axa ya Ti axb yb Ti should be noted that phase unwrapping for ambiguity resolution, þ½ðÞ; ; − ðÞ; ; nxa ya Ti nxb yb Ti a challenging task in conventional InSAR data processing, can be H. Tung, J.-C. Hu / Tectonophysics 578 (2012) 126–135 133

Table 2 Comparison of land subsidence revealed by precise leveling and SRD rate of PSInSAR at 11 benchmarks located in the study area.

Township Station name Elevation Elevation Elevation Leveling average subsidence PSInSAR subsidence Residual (1996/10) (1998/02) (1999/11) rate (cm/yr) rate along LOS (cm/yr) (cm/yr)

A Mailiao Fengan Elementary School 172.03 168.34 166.90 −1.67 −1.01 −0.66 B Mailiao Township Ofﬁce 578.52 571.37 567.41 −3.60 −2.54 −1.06 C Mailiao Work Station 574.11 570.92 −1.82 −2.38 0.56 D Lunbei Dayou Elementary School 1000.03 993.28 987.38 −3.45 −4.19 0.65 E Erhlun Youche Elementary School 1855.65 1855.56 1854.18 −0.48 −0.48 0 F Taishi Anshi Temple 111.36 106.63 −2.71 −3.52 0.81 G Sinsing Elementary School 233.23 228.60 225.66 −2.46 −1.40 −1.06 H Taishi Work Station 146.38 137.35 132.22 −4.59 −3.15 −1.44 I Baojhong Longyan Farm 997.54 988.33 982.79 −4.78 −4.27 −0.51 J Douliou Government of Yunlin County 4327.73 4327.91 4326.21 −0.49 −0.16 −0.33 K Yuanchang Yuanchang Cemetery 1019.59 1007.29 −7.03 −6.14 −0.89 L Jhongsiao Elementary School 980.22 972.99 965.79 −4.68 −4.55 −0.14 avoided through optimizing the objective function. In addition, the deformation estimation, there are 4566 points remained from the OC value is a good quality measure of the phase measurements at initial 5605 points according to the quality test. However, the linear the PS points. The larger an OC value is, the higher the accuracy of deformation measurement shows a similar pattern both with precise the estimated parameters is. Our study has shown that Δε and Δv leveling even the PS density is low (2.68 PS/km2). The slant range de- can be estimated accurately for a connection when its γ is greater formation rate of PSs relative to the benchmark at You-Che Elementary than 0.7. Therefore only connections with γ greater than 0.7 are School in Erhlun (see Table 2) can be observed in Fig. 6. The uplift rate used for analysis. up to 2 cm/yr and subsidence rate up to about 8 cm/yr are observed in the study area. Most areas with severe subsidence are located the 4. Results and discussion inner area of Yunlin County, especially in Baojhou, Tuku and Yuanchang Townships. However, the THSR is just located on the eastern Tuku 4.1. Results of land subsidence revealed by PSInSAR Township (Fig. 7). For comparison with the leveling data, the incident angles of radar were used to convert the LOS displacement to vertical We select the master image using Eq. (1), and then generate 32 deformation under the assumption of zero horizontal motion. This interferograms with the master image (1997/06/05). According to assumption has been carried out in previous land subsidence by InSAR the dense vegetation in the study area, the 32 interferograms reveal (e.g. Amelung et al., 1999; Galloway et al., 1998). The ten benchmarks less information of phase difference even the temporal baseline is of precise leveling (Table 2) have been chosen to constrain the vertical small; most of the interferograms are full of perturbation that we deformation rate from PSInSAR converted from SRD rate. The residual of are not able to observe any information of land subsidence. Based precise leveling and PS-InSAR is in the range from 0.81 to −1.44 cm/yr. on early repeat leveling results measured by ITRI, we chose You-Che In addition, westward horizontal displacements about 1 cm/yr Elementary School as the reference point with the minimum eleva- relative to reference point S01R (Fig. 1a) at stable continental margin tion change. After processing the data using the algorithm for linear do contribute to SRD rate with an elongation of 0.38 cm/yr along the

Fig. 6. The average rate of along light of sight (LOS) of PS points in the study area superimposed on 40 m DEM. Solid stars are the benchmarks of precise leveling chosen for constrains of PSInSAR results. The negative SRD (Slant Range displacement) rate represents land subsidence and positive ones represent uplift. 134 H. Tung, J.-C. Hu / Tectonophysics 578 (2012) 126–135

Fig. 7. The average vertical displacement map by PSs of the study area from 1996 to 1999. Solid rectangles are benchmarks of precise leveling. line of sight to satellite. However, recent study argument that about References half of the continuous GPS sites in the Los Angeles basin revealed superposed effects of tectonic motion and deformation associated Abidin, H.Z., Djaja, R., Darmawan, D., Hadi, S., Akbar, A., Rajiyowiryono, H., Sudibyo, Y.,Meilano,I.,Kasuma,M.A.,Kahar,J.,Subarya,C.,2001.Landsubsidenceof with fluid pumping (Bawden et al., 2001), thus the evaluation of Jakarta (Indonesia) and its geodetic monitoring system. Natural Hazards 23, the components of tectonic motion and deformation associated to 365–387. groundwater extraction are not taken into account in this study. Amelung, F., Galloway, D.L., Bell, J.W., Zebker, H.A., Laczniak, R.J., 1999. Sensing the ups and downs of Las Vegas: InSAR reveals structural control of land subsidence and aquifer-system deformation. Geology 27 (6), 483–486. Bawden, G.W., Thatcher, W., Stein, R.S., Hudnut, K.W., Peltzer, G., 2001. Tectonic con- 4.2. Long-term subsidence rate, shorten subsidence rate traction across Los Angeles after removal of groundwater pumping effects. Nature 412, 812–815. Berardino, P., Fornaro, G., Lanari, R., Sansosti, E., 2002. A new algorithm for surface From a tectonic point of view, most of the western coastal plains deformation monitoring based on small baseline differential SAR interferograms. are considered to be in the foreland basin of the Taiwan orogen, the IEEE Transactions on Geoscience and Remote Sensing 40 (11), 2375–2383. Buckley, S.M., Rosen, P.A., Hensley, S., Tapley, B.D., 2003. Land subsidence in Houston, newly initiated blind fault system in the widespread coastal plain. Texas, measured by radar interferometry and constrained by extensometers. Journal Liew et al. (2004) figured out that the marine incursion of the last of Geophysical Research 108 (B11), 2542. http://dx.doi.org/10.1029/2002JB001848. interglacial was recognized as between 100 and 150 m below surface Bürgmann, R., Rosen, P.A., Fielding, E.J., 2000. Synthetic aperture radar interferometry level which was widely distributed eastward all the way to the foot- to measure Earth's surface topography and its deformation. Annual Review of Earth and Planetary Sciences 28, 169–209. hills based on sedimentological and paleontological studies from CGS, 1999. The Investigation of Hydrogeology in the Choshui River Alluvial Fan, Taiwan cores in the central part of Choshui fan delta. The study area, under- (in Chinese). Central Geological Survey of Taiwan, Taipei. lain by the Mesozoic basement high (Fig. 1a), is the most stable area Chang, C.-C., Wang, T.-N., 2006. GPS monitoring ground subsidence associated with – seasonal underground water level decline: case analysis for a section of Taiwan with a very low subsidence rate of about 0.7 1.1 mm/yr since the High Speed Rail. Surveying and Land Information Science 66 (1), 45–54. last interglacial and a subsidence rate of 2–3 mm/yr since latest Chen, C.-T., Hu, J.-C., Lu, C.-Y., Lee, J.-C., Chan, Y.-C., 2007. Thirty-year land elevation change Pleistocene (Lai and Hsieh, 2003). Thus the long-term tectonic contri- from subsidence to uplift following the termination of groundwater pumping and its geological implications in the Metropolitan Taipei Basin, Northern Taiwan. Engineering bution is the minor component for the subsidence in the study area. Geology 95, 30–47. http://dx.doi.org/10.1016/j.enggeo.2007.09.001. Crosetto, M., Castillo, M., Arbiol, R., 2003. Urban subsidence monitoring using radar interferometry: Algorithms and validation. Photogrammetric Engineering and Re- mote Sensing 69 (7), 775–783. 5. Conclusion Ding, X.-L., Liu, G.-X., Li, Z.-W., Li, Z.-L., Chen, Y.-Q., 2004. Ground subsidence monitoring in Hong Kong with satellite SAR interferometry. Photogrammetric Engineering and Although the density of PS is low in Yunlin County due to flourishing Remote Sensing 70 (10), 1151–1156. vegetation. The PSInSAR result shows the serious subsidence areas Ding, X.-L., Li, Z.-W., Zhu, J.-J., Feng, G.-C., Long, J.-P., 2008. Atmospheric effects on InSAR measurements and their mitigation. Sensors 8, 5426–5448. located in Baojhou, Tuku and Yuanchang Townships and northern Ferretti, A., Prati, C., Rocca, F., 2000. Nonlinear subsidence rate estimation using perma- Mailiao, and the most subsidence rate is about 8 cm/yr in this time nent scatterers in differential SAR interferometry. IEEE Transactions on Geoscience – period in conjunction of Tuku and Yuanchang Townships, which is and Remote Sensing 38 (5), 2202 2212. Ferretti, A., Prati, C., Rocca, F., 2001. Permanent scatterers in SAR interferometry. IEEE quite reliable. This PSInSAR result has proven that we can effectively Transactions on Geoscience and Remote Sensing 39 (1), 8–20. reduce labor and cost by using this technique on observing land subsi- Galloway, D.L., Hudnut, K.W., Ingebritsen, S.E., Phillips, S.P., Peltzer, G., Rogez, F., Rosen, dence in Taiwan Island. In the future, the available L band ALOS data P.A., 1998. Detection of aquifer system compaction and land subsidence using interferometric synthetic aperture radar, Antelope Valley, Mojave Desert, California. and X band TerraSAR and COSMO-SkyMed radar satellites could be Water Resources Research 34 (10), 2573–2585. used to continue surveying land subsidence in Yunlin County. Hooper, A., 2006. Persistent scatterer radar interferometry for crustal deformation studies and modeling of volcanic deformation. Unpublished Ph. D. Dissertation, Institute of Geophysics, Stanford University, USA, 144 pp. Hooper, A., Zebker, H., Segall, P., Kampes, B., 2004. A new method for measuring defor- Acknowledgments mation on volcanoes and other natural terrains using InSAR persistent scatterers. Geophysical Research Letters 31 (23), L23611 (1–5). The comments and suggestions from Editor Fabrizio Storti, guest Hou, C.-S., Hu, J.-C., Shen, L.-C., Wang, J.-S., Chen, C.-L., Lai, T.-C., Huang, C., Yang, Y.-R., Chen, R.-F., Chen, Y.-G., Angelier, J., 2005. Estimation of subsidence using GPS editor Lionel Siame, Stephane Dominguez, M. Peyret and one anony- measurements, and related hazard: the Pingtung Plain, southwestern Taiwan. mous reviewer are deeply appreciated. This study was supported by Comptes Rendus Geoscience 337, 1184–1193. the National Science Council (project number: 97-2745-M-001-002). Hsieh, C.-S., Shih, T.-Y., Hu, J.-C., Tung, H., Huang, M.-H., Angelier, J., 2011. Using differential SAR interferometry to map land subsidence: a case study in the Pingtung We are grateful to the Water Resource Agency, Department of Economics, Plain of SW Taiwan. Natural Hazards 58 (3), 1331–1332. http://dx.doi.org/ Taiwan for the leveling data. 10.1007/s11069-011-9734-7. H. Tung, J.-C. Hu / Tectonophysics 578 (2012) 126–135 135

Hsu, S.-K., 1998. Plan for a ground water monitoring network in Taiwan. Hydrogeology Liu, C.-W., Lin, W.-S., Shang, C., Liu, S.-H., 2001. The effect of clay dehydration on land sub- Journal 6, 405–415. sidence in the Yun-Lin coastal area, Taiwan. Environmental Geology 40 (4–5), 518–527. Hu, J.-C., Yu, S.-B., Angelier, J., Chu, H.-T., 2001. Active deformation of Taiwan from GPS Liu, C.-H., Pan, Y.-W., Liao, J.-J., Huang, C.-T., Ouyang, S., 2004. Characterization of land subsi- measurements and numerical simulations. Journal of Geophysical Research 106, dence in the Chohui River alluvial fan, Taiwan. Environmental Geology 45, 1154–1166. 2265–2280. Liu, G., Buckley, S.M., Ding, X., Chen, Q., Luo, X., 2009. Estimating spatiotemporal Huang, C.-Y., Yuan, P.B., Tsao, S.-J., 2006a. Temporal and spatial records of active arc- ground deformation with improved permanent-scatterer radar interferometry. continent collision in Taiwan: a synthesis. Geological Society of America Bulletin IEEE Transactions on Geosciences and Remote Sensing 47 (9), 3209–3219. http:// 118 (3–4), 274–288. http://dx.doi.org/10.1130/B25527.1. dx.doi.org/10.1109/TGRS.2009.2028797. Huang, M.-H., Hu, J.-C., Hsieh, C.-S., Ching, K.-E., Rau, R.-J., Pathier, E., 2006b. A growing Liu, G.-X., Luo, X.-J., Chen, Q., Huang, D.-F., Ding, X.-L., 2008. Detecting land subsidence structure near the deformation front in SW Taiwan deduced from SAR interferom- in Shanghai by PS-networking SAR interferometry. Sensors 8, 4725–4741. etry and geodetic observation. Geophysical Research Letters 33, L12305. http:// Massonnet, D., Feigl, K.L., 1998. Radar interferometry and its application to changes in dx.doi.org/10.1029/2005GL025613. the Earth's surface. Reviews of Geophysics 36 (4), 441–500. Huang, M.-H., Hu, J.-C., Ching, K.-E., Rau, R.-J., Hsieh, C.-S., Pathier, E., Fruneau, B., Mora, O., Mallorqui, J.J., Broquetas, A., 2003. Linear and nonlinear terrain deformation Deffontaines, B., 2009. Active deformation of Tainan Tableland of southwestern maps from a reduced set of interferometric SAR images. IEEE Transactions on Taiwan based on geodetic measurements and SAR interferometry. Tectonophysics Geoscience and Remote Sensing 41 (10), 2243–2253. 466, 322–344. http://dx.doi.org/10.1016/j.tecto.2007.11.020. Motagh, M., Djamour, Y., Walter, T.R., Wetzel, H.-U., Zschau, J., Arabi, S., 2007. Land sub- Hung, W.-C., Liu, C.-S., 2007. Taiwan land subsidence monitoring and surveying analysis. sidence in Mashhad Valley, northeast Iran: results from InSAR, levelling and GPS. Report of Industrial Technology Research Institute (ITRI), Hsinchu. 335 pp. (in Geophysical Journal International 168 (2), 518–526. http://dx.doi.org/10.1111/ Chinese). j.1365-246X.2006.03246.x. Hung, W.-C., Hwang, C., Chang, C.-P., Yen, J.-Y., Liu, C.-H., Yang, W.-H., 2010. Monitoring Osmanoglu, B., Dixon, T.H., Wdowinski, S., Enrique, C.-C., Jiang, Y., 2010. Mexico City sub- severe aquifer-system compaction and land subsidence in Taiwan using multiple sidence observed with Persistent Scatterer InSAR. International Journal of Applied sensors: Yunlin, the southern Choshui River Alluvial Fan. Environmental Earth Earth Observation and Geoinformation. http://dx.doi.org/10.1016/j.jag.2010.05.009. Sciences 59 (7), 1535–1548. http://dx.doi.org/10.1007/s12665-009-0139-9. Phien-wej, N., Giao, P.H., Nutalaya, P., 2006. Land subsidence in Bangkok, Thailand. Hung, W.-C., Hwang, C., Chen, Y.-A., Chang, C.-P., Yen, J.-Y., Hooper, A., Yang, C.-Y., 2011. Engineering Geology 82, 187–201. Surface deformation from persistent scatterers SAR interferometry and fusion with Pritchard, M.E., Simons, M., 2002. A satellite geodetic survey of large-scale deformation leveling data: a case study over the Choshui River Alluvial Fan, Taiwan. Remote of volcanic centres in the central Andes. Nature 418, 167–171. Sensing of Environment 115, 957–967. Tang, C.-P., Tang, S.-Y., 2006. Democratization and capacity building for environmental Hwang, C., Hung, W.-C., Liu, C.-H., 2008. Results of geodetic and geotechnical monitor- governance: managing land subsidence in Taiwan. Environment and Planning A ing of subsidence for Taiwan High Speed Rail operation. Natural Hazards 47, 1–16. 38, 1131–1147. http://dx.doi.org/10.1068/a37375. http://dx.doi.org/10.1007/s11069-007-92115. Teatini, P., Ferronato, M., Gambolati, G., Bertoni, W., Gonella, M., 2005. A century of land Kampes, B.M., 2005. Radar Interferometry: Persistent Scatterer Technique. Springer, subsidence in Ravenna, Italy. Environmental Geology 47, 831–846. Germany. Wang, C.-T., Chen, K.-S., Hu, J.-C., Chang, W.-Y., Boerner, W.M., 2011. Mapping land uplift Kampes, B.M., Hanssen, R.F., 2004. Ambiguity resolution for permanent scatterer interfer- and subsidence in the Industrial Parks in northern Taiwan by radar interferometry. ometry. IEEE Transactions on Geoscience and Remote Sensing 42 (11), 2446–2453. International Journal of Remote Sensing 32 (21), 6527–6538. http://dx.doi.org/ Lai, T.-H., Hsieh, M.-L., 2003. Late Quaternary vertical rock-movement rates of the 10.1080/01431161.2010.512932. coastal plains of Taiwan. Ann. Meeting Geol. Soc. China, Programs with Abstracts, Wright, T.J., 2002. Remote monitoring of the earthquake cycle using satellite radar p. 119. interferometry. Philosophical Transactions of the Royal Society of London, Series Liew, P.-M., Hsieh, M.-L., Shyu, B.H., 2004. An overview of coastal development in a A 360, 2873–2888. http://dx.doi.org/10.1098/rsta.2002.1094. Young Mountain Belt-Taiwan. Quaternary International 115–116, 39–45. Yen, J.-Y., Chen, K.-S., Chang, C.-P., Boerner, W.-M., 2008. Evaluation of earthquake Lin, A.T., Watts, A.B., 2002. Origin of the West Taiwan basin by orogenic loading and potential and surface deformation by differential interferometry. Remote Sensing ﬂexure of a rifted continental margin. Journal of Geophysical Research 107 (B9), of Environment 112, 782–795. http://dx.doi.org/10.1016/j.res.2007.06.012. 2185. http://dx.doi.org/10.1029/2001JB000669. Zebker, H.A., Villaseno, J., 1992. Decorrelation in interferometric radar echoes. IEEE Lin, K.-C., Hu, J.-C., Ching, K.-E., Angelier, J., Rau, R.-J., Yu, S.-B., Tsai, C.-H., Shin, T.-C., Transactions on Geoscience and Remote Sensing 30 (5), 950–959. Huang, M.-H., 2010. GPS crustal deformation, strain rate and seismic activity Zhang, Z.C., 2005. Plan our country water resources policy: take Yunlin County as the after the 1999 Chi-Chi earthquake in Taiwan. Journal of Geophysical Research example (in Chinese, with English abstract). Unpublished Master Thesis, National 115, B07404. http://dx.doi.org/10.1029/2009JB006417. Sun Yat-sen University, Kaohsiung, 92 pp.