Deformation Monitoring and Evaluation of Mountain Slope Stability Combined With Ground-based Radar and Spaceborne InSAR Methods

8 In this paper, ground-based radar and spaceborne Interferometric Synthetic 9 Aperture Radar (InSAR) images were combined to monitor slope stability and 10 analyze the main deformation factors of an ancient landslide on the right bank of the 11 Dajinchuan River in Danba County, Sichuan Province, China. We applied the short 12 baseline set (SBAS) time series strategy with 656 scenes of ground-based radar 13 between September 13 - 17, 2019, and 62 scenes of Sentinel - 1 data from July 2018 to 14 October 2020. Combined with high-resolution satellite images and digital elevation 15 model (DEM) data, we acquired trace and quantitative deformation features and 16 discussed the factors that contributed to slope instability, such as geological structure, 17 topography, external environment and human activities. The largest deformation area 18 detected by ground-based radar is located in the bedrock above the target area with a 19 maximum cumulative deformation of more than 30 mm during the detection time. The 20 maximum average annual deformation rate detected over the region by spaceborne 21 InSAR is over 40 mm/a. We analyzed the differences between the ground-based radar 22 and spaceborne InSAR and the reasons for the differences. This study provides 23 references and suggestions for investigating potential landslide risks by combining 24 ground-based radar and spaceborne InSAR technology. 25

China, where the terrain is extremely complex and significantly characterized by an 116 alpine canyon landform. The Dadu River cuts across Danba County from north to south 117 (Fig 1), with river channels of 1000 to 3000 m in depth. In contrast, influenced by the 118 high mountain and canyon topography, Danba County is characterized by the Tibetan 119 Plateau monsoon. Moreover, the area is also characterized by obvious vertical zonality 120 due to topographic factors. With increasing altitude, both the temperature and 121 evaporation decrease gradually; however, the precipitation increases. The precipitation 122 is mainly concentrated between May and September, the rainy season and dry season 123 are obviously different, and the temperature varies greatly from day to night. The study area of this paper is located on the right bank of the Dajinchuan River 136 (Fig 2), a tributary of the Dadu River. The study area is also situated at the foot of the 137 right front slope of the old, large Jiaju landslide group, which consists of many 138 secondary landslides that led to multiple bank deformation events along the 139 Dajinchuan River. The observation area of the ground-based radar in this study is the 140 slope bedrock of a secondary ancient landslide. The front edge of the ancient landslide 141 is approximately 1.4 km in length, and the height difference between the front and back 142 edges of the landslide is more than 700 meters. The general terrain slope exceeds 25 143 degrees according to the DEM data. Two rock fractures extending in the northeast 144 direction can be clearly interpreted from the Google Earth images in the observation 145 area of the ground-based radar (Fig 2(c)). The upper fracture is approximately 820 m in 146 length, and the lower fracture is approximately 600 m in length. To study the 147 deformation feature of this bedrock slope of the ancient landslide, we utilized 148 ground-based radar and the satellite InSAR method. The deformation results measured 149 by the two methods are analyzed, and these results may help to better understand the 150 landslide in similar studies.

Measurement principle of ground-based radar 162
The ground-based radar data used in this study are acquired by the portable radar 163 interferometer GPRI-II (Gamma Portable Radar Interferometer) produced by GAMMA 164 company. This interferometer uses real aperture radar with a radar frequency of 17.2 165 GHz (Ku band), bandwidth of 200 MHz, wavelength of 0.0176 m and effective 166 measurement range from 0.05 to 10 km. The deformation monitoring accuracy can 167 reach the submillimeter level with an azimuth resolution of 6.28 m (when the distance 168 is 1 km) and a range resolution of 0.75 m. The GPRI-II is mounted on a tripod and 169 measured at 360° by a rotating scanner with three antennas, one transmitting signal and 170 two receiving the echo signal. 171 GPRI-II uses a frequency-modulated continuous wave (FMCW) to measure the 172 velocity and distance of the target by the frequency differences between the transmitted 173 signal and received signal. This technique is suitable for data acquisition and digital 174 signal processing with low difference frequency signals and simple hardware 175 processing. Compared with SFCW (step frequency continuous wave), GPRI-II can 176 improve the scanning speed and reduce the influence of the atmospheric delay phase on 177 the monitoring precision, reducing the phase distortion caused by system noise in the 178 long-term scanning process. At the same time, GPRI-II can generate a DEM by means 179 of two antennas. The system generally uses the continuous observation mode to 180 continuously observe the target area. 181 The range resolution of GPRI-II is 182 where C is the speed of light and B is the bandwidth. As seen from the above 184 equation, the range resolution is independent of the distance between the instrument 185 and the observed target. 186 The azimuth resolution is 187 θ−3dB is the width of the half-power wave velocity, and R is the azimuth distance. 189 Since the spatial baseline of GPRI-II is 0 and the observation mode is continuous, 190 the interferometric phase of GPRI-II does not include the terrain phase and the 191 geometric phase component between the two positions; thus, the interferometric phase 192 is 193 In the formula, φ is the deformation phase, φ is the atmospheric delay 195 phase, and φ is the noise phase. 196 Compared with the spaceborne InSAR system, the ground-based radar system 197 has several unique advantages (Wu et al. 2019; Tiandong Chen 2020). First, the 198 precision of the ground-based radar is higher up to the submillimeter level as its 199 wavelength is shorter. Second, the observation period of ground-based radar is shorter, 200 and the time sampling rate is higher to simplify the phase unwrapping process and 201 achieve rapid real-time monitoring. Third, the ground-based radar system is more  area. Second, a reasonable observation distance, that is, the appropriate monitoring 213 distance, should be selected according to the actual situation of the site, such as the 214 topography, engineering, hydrology and other conditions. The larger the distance is, the 215 weaker the radar receiving echo signal, and the worse the monitoring effect. The third 216 condition is that the equipment should be placed on a stable observation platform to 217 reduce the influence of any small equipment movement on the observation accuracy. 218 The last condition is a suitable viewing angle. The smaller the angle is, the more 219 sensitive the radar is to the intensity of the deformation signal, but it is disadvantageous 220 to receive the echo signal. In ground-based radar deformation monitoring, the 221 observation parameters should be adjusted at the beginning according to the 222 environmental factors and the quality of the observation data because the subsequent 223 data processing accuracy has a great impact.  The time interval between the two receiving antennas was approximately 10 minutes 228 and one scene, and the total number of images was 656 scenes. The observed 229 parameters are shown in Table 1. 230  results. The methods of coherence point extraction mainly include the amplitude 243 departure threshold method, local coherence method (coherence coefficient threshold 244 method) and nonlocal method. In this paper, a nonlocal method is used to extract 245 coherent points by selecting homogeneous or similar pixel estimates from the 246 surroundings of each resolution unit. 247 In the process of phase unwrapping, there is a 2kπ relation between the initial 248 phase and the true phase in the interferogram. The initial phase is the winding phase 249 between -π and π, which is the main value of the true phase. The interference phase 250 needs to be decoded to obtain the true phase. Considering the stability and time effect 251 of phase unwrapping, the three-dimensional phase unwrapping method is adopted in 252 this paper. 253 Atmospheric correction is also necessary for ground-based radar data (Xining 254 Zhang et al. 2017). Ground-based radar relies on the phase information of radar signals 255 for ranging, but the accuracy of ranging is affected by the changes in the refractive 256 index of the radar signals because of the atmosphere. Even for short-term monitoring, 257 shortwave band ground-based radar is also very sensitive to weather changes, so 258 improving the measurement error caused by atmospheric phases has become the key 259 technology to improve the observation accuracy of ground-based radar. In this paper, an 260 iterative decomposition model is used to correct the effect of atmospheric variations on 261 the deformation results. 262 We use the singular value decomposition method (Li et al. 2013) to generate the 263 deformation time series diagram. The deformation characteristics, including the spatial 264 distribution, deformation intensity and future development trend, can be estimated by 265 the time series map of deformation, which provides a basis for emergency response. 266

Other auxiliary data 267
To investigate the features of the slope, as well as for comparison with the 268 deformation results obtained by ground-based radar GPRI-II, remote sensing data from 269 different platforms were also employed, including spaceborne InSAR and 270 high-resolution Google Earth images. A total of 62 scenes of Sentinel-1 data from July 271 2018 to October 2020 were selected for deformation analysis. 272 Other auxiliary data, including geological maps and meteorological and 273 hydrological data, were also employed in the study. 274

Deformation results of ground-based radar 276
The effective monitoring time of ground-based radar is from 8:27 on 13 277 September to 21:37 on 13 September and from 16:46 on 15 September to 13:26 on 17 278 September. The maximum monitoring distance is approximately 1300 m, and the 279 minimum monitoring distance is approximately 450 m. According to the processing 280 flow in section 3.3, we set the parameters, which include the unit window size and time 281 baseline. Since the data format of the data acquisition system synchronized to the 282 ground-based radar data processing system is binary, it needs to be converted to MAT 283 format. The average coherence criterion is used to extract the coherent points, and the 284 nonlocal coherence algorithm is used to set the coherence threshold value to 0.35, the 285 nonlocal window to 15, the similarity threshold value to 0.9, the minimum similarity 286 point to 10, and the maximum similarity point to 45. Some of the differential 287 interferograms produced during ground-based radar data processing are shown in

Deformation results from spaceborne InSAR 309
Sixty-two scenes of Sentinel-I downorbiting data covering the study area from 310 July 2018 to October 2020 were processed using GAMMA software, and the average 311 annual deformation rate of the study area over this period was obtained using the short 312 baseline set (SBAS) processing method (Fig 5(a)

Deformation analysis of the mountain slope 329
We match the ground-based radar and spaceborne InSAR data accurately and 330 superpose to the Google Earth 3D image (Fig 6) to compare the image features of 331 different deformation regions. According to the interpretation of satellite remote 332 sensing images, the main target area of this observation is located on the right wall of a 333 pre-existing ancient landslide. The accumulation body of the early ancient landslide is 334 mainly located at the left foot of the ancient landslide, which is the upper part of the 335 scenic area. Some accumulation bodies remain on the right side of the ancient landslide. 336 The target area observed by ground-based radar is mainly on the right side of the 337 ancient landslide, and the back wall is the blind area, which is not effectively covered.

Main factors influencing slope deformation 360
In general, the factors that affect the stability and deformation of geological slopes 361 include geological structure, topography, external environment and human activities. In terms of topography and geomorphology, the slope angle of a geological body 384 is an important factor that affects its stability. Generally, geological bodies with slope 385 angles greater than 10 are subject to unstable deformation under the action of gravity 386 (Donnarumma et al. 2013;Luo et al. 2020). According to the regional DEM, the slope 387 angle of the whole terrain is more than 30° and the gradient is more than 0.58, while the 388 slope angle of the local slope of ground-based radar is more than 20° and the gradient is 389 more than 0.41 (Fig 7). Whether considering the whole terrain or the local terrain, the ground-based radar is located at the foot of the slope, so the observed deformation is 462 closer to the real deformation characteristics of the landslide. The spaceborne InSAR 463 data reference point is relatively stable in the whole image, which is located a large 464 distance from the slope, so the deformation shown is also relative to other reference 465 points. There may be overall deformation in a region that either increases or counteracts 466 the true deformation of the slope. 467 In terms of imaging resolution and deformation monitoring accuracy, the medium 468 resolution of the spaceborne InSAR Sentinel-1 is 5 m × 20 m, and the time series 469 processing accuracy is at the mm level. The range resolution of the ground-based radar 470 is 0.75 m, and the azimuth resolution at 1 km is 6.8 mrad with submillimeter accuracy. 471 The data precision of the ground-based radar is obviously higher than that of 472 spaceborne data, but it is not suitable for long-term and large-scale observations, and 473 spaceborne data are not suitable for small-scale deformation objects. 474 In data processing, the space baseline of ground-based radar is 0, there is no need 475 for terrain phase compensation, and there is no influence from an orbit error. The 476 influence of the atmosphere is weak due to the limited observation distance. In contrast, 477 there is not only the influence of the spatial baseline but also a large interference of 478 atmospheric error in spaceborne InSAR data processing, so it is necessary to repeatedly 479 remove the influence of internal and external noise from various systems, which will 480 greatly reduce the accuracy of the observation data. The above studies show that there are usually some differences between the 484 ground-based radar and spaceborne InSAR results of deformation monitoring. These  Table 2 for different scenarios and stages. 531     presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 3
Examples of differential interferograms generated during the processing of ground-based radar data. The two numbers in the upper part of each gure denote the observation dates of two data points used to generate the differential interferogram. 20190913084754-20190913085754 denotes the differential interferogram generated between two data points observed    (a) Regional annual deformation rate map measured by Sentinel-1 data, and P1, P2 and P3 are the three points selected in the observation area of ground-based radar. The gray line is the scanning angular scope of ground-based radar. (b) The line-of-sight cumulative deformation map of three selected points P1, P2 and P3 in the study area. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.   Sketch map of the slope and angle of the study area Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.