5.1. Comparison of the InSAR-Derived Results with the Leveling Data
To assess the accuracy of the InSAR-derived results over the three observation periods, a quantitative comparison of the time-series displacements with the leveling survey data provided by the Chiba Prefecture at 22 measurement points was performed. To locate each leveling point, we referenced the online version of the Chiba information map and the illustration figures of each leveling point provided [52
]. For the InSAR measurement points, especially those in incoherent areas, the pixels that lay within 100 m of the corresponding leveling points were assigned, and the average velocity of these pixels was calculated. We selected the leveling data in the same overlapping periods as the three InSAR measurement periods. We assumed the horizontal deformation was negligible, and the LOS displacement velocity was converted into the vertical displacement velocity by dividing the cosine of the sensor incidence angle [53
shows the spatial distribution of the leveling points and the comparison between the leveling and InSAR-derived linear subsidence rate. Note that the number of leveling points are different among the different InSAR observation periods; 17 leveling points were used for the comparison of the ERS-1/-2 and PALSAR observation periods, while 21 leveling points were used for the PALSAR-2 observation period, which is due to five new leveling points being established after the 2011 Tohoku Earthquake and the leveling point U-12A being missing in 2016. We also used different plot scales (20 mm/year vs. 12 mm/year) and (2/4 mm vs. 1/2 mm for error lines), due to the smaller errors shown in PALSAR data using the SBAS method (Figure 12
d). The comparison results show that the results from the ERS-1/-2 data using the SBAS method have the largest root mean square errors (RMSEs) of 4.4 mm/year, while the results from PALSAR and PALSAR-2 data using the SBAS method have the smallest RMSEs of 0.9 and 2.2 mm/year, respectively. For the ERS-1/-2 and PALSAR data, more than 12 out of the 17 measurement points showed a residual value of less than 4 mm/year (Figure 12
a–d); for the PALSAR-2 data, and 14 out of the 21 measurement points showed a residual value of less than 2 mm/year (Figure 12
e). As shown in Figure 12
a,b,e, the results from the ERS-1/-2 and PALSAR-2 data using the PSI and SBAS method showed the largest discrepancies at several leveling points. This may have been caused by the low coherence of ERS-1/-2 datasets and the contribution of phase noise. The fewer PALSAR-2 image pairs and the sudden elevation changes in the ground, i.e., the leveling point U-17 subsided by the influence of construction work during 2015–2016 [37
], may also affect the comparison result. Nevertheless, according to these comparisons, the InSAR-derived results agree relatively well with the result of the leveling measurements and suggest the reliability of the InSAR-measured subsidence rate.
5.2. Spatial and Temporal Patterns of Land Subsidence
To further reveal the land subsidence patterns in different districts over the three observation periods, we generated the spatial distribution map of difference of land subsidence rates (Figure 13
) using the ArcGIS 10.3 (Esri, Redlands, CA, USA) spatial analyst tool. As the incidence angles of those sensors are different, before comparison, the LOS displacement velocity was converted into the vertical displacement velocity by dividing the cosine of the sensor incidence angle [53
]. It can be seen from Figure 13
a that the areas in the central town (i.e., Maihama (A), Tekkodori, Benten, Imagawa (B) and Irifune (C)) and new town (i.e., Takasu (E), Minato (F) and Chidori (G)) show slight to moderate subsidence with a 2–13 mm/year rate during the ERS-1/-2 observation period. From Figure 13
b, we can see that the subsidence rate in some of the districts of the central town (e.g., Benten, Tekkodori and Imagawa (B)) has decreased up to 12 mm/year; while the areas in the new town showed increasing subsidence up to 28 mm/year, especially in Hinode (D) and Chidori (G). The comparison of PALSAR-2 and PALSAR estimated subsidence rate show that, the previous subsiding areas were experiencing a reduced subsiding rate, except some localized subsidence in the new town (Figure 13
c). The comparison of PALSAR-2 and ERS-1/-2 estimated subsidence rate show that, the subsidence in both of the central town and new town has significantly decreased, except for areas in Maihama (A), Irifune (C), Hinode and Akemi (D) (Figure 13
d). In general, most of those areas in the central town are residential and commercial amusement land, while the bew town are parks and industrial land. The subsidence in parks can only be caused by the natural soil consolidation, while in the residential, commercial and industrial areas, the subsidence may be caused from the integrated effect of numerous natural and anthropogenic processes.
5.3. The Use of Different SAR Sensors in Land Subsidence Monitoring
The number of satellite data sources is currently increasing steadily. These datasets from the previous SAR sensors such as ESA archive (ERS-1/-2, ENVISAT) as well as the new generation of C, X and L-band SAR images provided by the RADARSAT-2, Sentinel-1A, ALOS-2, TerraSAR-X, Tandem-X and the COSMO-SkyMed constellation, etc. have enabled us to compute the time series of the occurred and on-going surface displacements from regional scale to individual buildings. In particular, the exploitation of the free and open access data archives collected by the Sentinel-1A system permit us to conduct continuous land deformation analysis over large areas.
In this study, three different SAR datasets, the ERS-1/-2, ALOS PALSAR and ALOS-2 PALSAR-2, were used to monitor the long-term land subsidence in Urayasu City. The C-band has a shorter wavelength and hence better displacement sensitivity, and the L-band has longer wavelength and lower frequency showing more extensive coverage over natural areas and less temporal decorrelation. The data acquired by these satellites cover long periods of time and enabled us to perform long-term deformation monitoring of the study area. However, those different sensors have different imaging parameters, e.g., spatial and temporal resolution, incidence angle, and wavelength, which show different characteristics in terms of their maximum detection gradient, degree of decorrelation, capability of noise rejection, etc. The different imaging parameters and the use of an uneven number of images among different sensors cause some difficulties in comparing their performance and the quality control of multi-sensor InSAR results. Moreover, the low resolution and the longer revisit time of ERS-1/-2 and PALSAR has prevented us from observing short-term land deformations caused by the anthropogenic activities. Furthermore, data gaps between the PALSAR and PALSAR-2 caused some difficulties in analyzing InSAR results.
5.4. Land Subsidence and Possible Causes
The origin of land subsidence in coastal areas can be summarized into two categories: either caused by natural causes (e.g., natural compaction/consolidation of soil or tectonic movements, such as earthquakes) or anthropogenic activities (e.g., oil, gas and ground water exploitation). In some cases, the pattern of land subsidence might be even more complicated when it is caused by the combined effects of multiple factors at different scales. In Urayasu City, since most of the areas are land-filled, the natural consolidation of soil is postulated to be the primary driver of land subsidence. To further analyze the relation between subsidence and soil geology, we compared the InSAR-derived subsidence areas with the geologic map showing the depth of the upper surface of the solid geological stratum in Urayasu City (Figure 14
a) and found a remarkable spatial correlation between the geologic map of the soil properties and the subsiding areas. In most of the reclaimed zones, the upper layer of soil filled with hill sand and dredged sandy soil (FS) with a standard penetration test (SPT) N-value of 2–8; an alluvial sand layer (AS) with SPT N-value of 10–20 underlies the filled layer; a very soft alluvial clay (AC) is deposited under the AS layer with a low SPT N-value of 0–5; a diluvial (Pleistocene) dense sand layer (DS) with SPT N-value of 50 or greater is deposited blow the AC layer (Figure 14
]. Along the line A–A’, the thickness of AC layer increases significantly between the Naka-Machi and Shin-Machi area, and it continues towards the sea (Figure 14
b). As the consolidation of soil occurs in soft clay deposits, the thick AC layers in Naka-Machi and Shin-Machi area are most probably responsible for the continuing subsidence in Urayasu City.
As shown in Figure 14
a, the depth of the bottom of the alluvial layers increases from 20 m in Moto-Machi to about 40 m in Shin-Machi, with several narrow-buried valleys of up to 70 m in depth. The buried-valleys, which are about 60 m deep, exist directly below the Minato, Chidori, Tekkodori, Imagawa, Akemi and Irifune areas, causing complicated changes in the thickness of the soft ground in those areas, while the depth increases up to 80 m in Maihama where the largest subsidence occurred. This further suggests that the areas undergoing large subsidence correspond to those having thick layers of soft soil over a stiff basement. The Moto-Machi area, with soil deposits consisting of sandy soils with an alluvial origin, was quite stable over the observation period, while the Naka-Machi and Shin-Machi areas, with thick layers of fine-grained soft soil overlying a stiff basement, had significant land subsidence over the study period. However, considering the complexity of the land use and the anthropogenic activities in different districts of Urayasu City, the subsidence may not be solely caused by the natural consolidation, but also from the integrated effects of numerous natural and anthropogenic processes.
The additional load of buildings and structures is also considered to be one of the causes of land subsidence in urban areas [56
]. In Urayasu City, since the establishment of Urayasu town in 1909 in the old town (Moto-Machi)—a naturally formed Holocene lowland—the natural soil consolidation might be gradually reduced and stopped. Besides this, the density of buildings in the old town are lower than the central town. Many houses, commercial buildings and public facilities were built in the central town during the first phase of the project, ending in 1975. Meanwhile, many high-rise buildings, universities, hotels and storehouses were built in the new town during the second phase of the project, ending in 1980 [22
]. The additional load during and after the building construction, especially the high-rise buildings, could transfer a high loading to the ground and may eventually lead to substantial land subsidence. However, these buildings use a pile foundation to satisfy bearing capacity and deformation and may not show significant subsidence while the surrounding areas are subsiding. Figure 15
shows the InSAR-derived subsidence velocity (2006–2010) and the locations of high-rise buildings. Most of those buildings show stability, whereas their surroundings show land subsidence. However, further investigations are expected to determine the relationship between land subsidence and the building density/high rise buildings.
The Maihama district in the central town, where Tokyo Disneyland is located, showed significant subsidence throughout the whole InSAR observation period. However, in this area, the pattern of land subsidence may be even more complicated due to the continuous construction and renovation/redevelopment of the fantasy-land and other anthropogenic activities. The SAR images with a low resolution and longer revisiting time, and the linear inversion model used in the InSAR processing, may hinder the effective monitoring of short-term movements such as those induced by human activities and may cause some biased results. Therefore, more high-resolution SAR data with a short revisiting time and further investigation is required to understand the intricacies of the relationship between land subsidence, natural consolidation and load of buildings.
Ground water exploitation is one of the major causes of land subsidence in many coastal cities, such as in Jakarta [5
], Bangkok [8
] and Shanghai [57
]. Nevertheless, this may not be the cause of land subsidence in Urayasu City; this is because the ground water exploitation was gradually reduced and stopped in 1993 [58
], and the city receives water from a water purification plant which uses the main water sources of the Tone river and Edogawa river [59
]. Moreover, since April 1992, Chiba Prefecture has been implementing restriction rules on groundwater use for the highly susceptible areas of land subsidence, including Urayasu City [60
]. Thus, the ground water exploitation has insignificant impacts on land subsidence in Urayasu City.
As an earthquake-prone country, earthquakes happen frequently in Japan. Earthquakes have significant influences on coastal areas, especially on reclaimed land. In the 2011 Tohoku earthquake, houses and infrastructures were severely damaged due to soil liquefaction in Urayasu City [22
]. In addition, long-term ground settlement was also observed after the earthquake, and the degree of subsidence was different in areas where reclaimed soils were improved or not [24
]. In the areas where the soil was not improved, the subsidence may have been accelerated by the earthquake. The InSAR observation results derived from the PALSAR-2 data showed significant continuing land subsidence near the levee areas (mostly parks and vacant lands), which may have been accelerated by the effects of the earthquake. However, most areas showed a decrease of land subsidence, this may be related to the fact that the PALSAR-2 observations (December 2014 to November 2017) were collected almost 4 years after the Tohoku earthquake, and considering the soil aging effect and soil improvement, the land settlement in most of those areas caused by the natural soil consolidation and the earthquake might gradually decrease. It is worth mentioning that after the earthquake, the Urayasu government started to test several countermeasure methods, such as lowering the ground water level and grid wall soil improvement. Finally, Urayasu has adopted the grid wall soil improvements as a countermeasure to prevent future risks [61
]. This project may also have played a positive role in alleviating the land subsidence in Urayasu City.