Figures & data
Figure 1. The location of the study area. (a) Location of Beijing; (b) Geography of Beijing Plain; (c) Location of SAR datasets.
![Figure 1. The location of the study area. (a) Location of Beijing; (b) Geography of Beijing Plain; (c) Location of SAR datasets.](/cms/asset/c6f1f488-5374-4402-b9ad-8629a85687b8/tgrs_a_2230689_f0001_oc.jpg)
Table 1. Main information of the Envisat ASAR and Radarsat-2.
Figure 3. Flowchart of identification of NoData grids. The yellow grids have no PS points and time series deformation need to be reconstructed.
![Figure 3. Flowchart of identification of NoData grids. The yellow grids have no PS points and time series deformation need to be reconstructed.](/cms/asset/b9d3b427-f8d7-49a1-8794-8b11443fc3b3/tgrs_a_2230689_f0003_oc.jpg)
Figure 4. Deformation rate derived from (a) Envisat ASAR (2003–2010), (b) Radarsat-2 (2010–2016) and (c) Radarsat-2 (2017–2020); (d) time series fusion during 2003–2020.
![Figure 4. Deformation rate derived from (a) Envisat ASAR (2003–2010), (b) Radarsat-2 (2010–2016) and (c) Radarsat-2 (2017–2020); (d) time series fusion during 2003–2020.](/cms/asset/142083c5-54bd-4077-924d-7dcb4decb9da/tgrs_a_2230689_f0004_oc.jpg)
Figure 5. Comparison of deformation derived from PS-InSAR and levelling observations. (a) Scatter plot of deformation derived from PS-InSAR and levelling observations during 2006–2013 and 2015–2016. (b, c) Time series deformation during 2003–2020.
![Figure 5. Comparison of deformation derived from PS-InSAR and levelling observations. (a) Scatter plot of deformation derived from PS-InSAR and levelling observations during 2006–2013 and 2015–2016. (b, c) Time series deformation during 2003–2020.](/cms/asset/f94964a3-8393-43fc-ba8f-c29b916309ca/tgrs_a_2230689_f0005_b.gif)
Figure 6. (a) R2 and (b) RMSE of fitted cumulative deformation derived from linear, quadratic, cubic and quartic fitting.
![Figure 6. (a) R2 and (b) RMSE of fitted cumulative deformation derived from linear, quadratic, cubic and quartic fitting.](/cms/asset/2fae4ec0-2199-4c3b-911a-bc7087d056c4/tgrs_a_2230689_f0006_oc.jpg)
Figure 7. Time-series cumulative deformation modelled by cubic polynomial fitting curve for (a) Point A and (b) Point B.
![Figure 7. Time-series cumulative deformation modelled by cubic polynomial fitting curve for (a) Point A and (b) Point B.](/cms/asset/1c889746-408a-405a-8a54-9e1d9e73b81f/tgrs_a_2230689_f0007_oc.jpg)
Table 2. Points at different scale grids.
Figure 10. (a) Spatial distribution of five clusters clustered by ISODATA; (b) R2 and (c) RMSE of the estimated ,
,
, and
.
![Figure 10. (a) Spatial distribution of five clusters clustered by ISODATA; (b) R2 and (c) RMSE of the estimated an0, an1, an2, and an3.](/cms/asset/83d598d7-7a60-44ff-8395-b591b3a032ce/tgrs_a_2230689_f0010_oc.jpg)
Figure 11. (a) RMSE and MAPE of reconstructed deformation of test samples with different deformation rate. (b) Scatterplots of cumulative deformation the improved MLS-SVR against that derived from PS-InSAR from 2003–2020. (c) and (d) Cumulative deformation time series reconstructed by MLS-SVR and derived from PS-InSAR of two test samples. (e) and (f) Cumulative deformation time series reconstructed by MLS-SVR and measured from levelling measurements.
![Figure 11. (a) RMSE and MAPE of reconstructed deformation of test samples with different deformation rate. (b) Scatterplots of cumulative deformation the improved MLS-SVR against that derived from PS-InSAR from 2003–2020. (c) and (d) Cumulative deformation time series reconstructed by MLS-SVR and derived from PS-InSAR of two test samples. (e) and (f) Cumulative deformation time series reconstructed by MLS-SVR and measured from levelling measurements.](/cms/asset/8e4caa71-a88b-4034-aa5c-12ebdadecfc9/tgrs_a_2230689_f0011_oc.jpg)
Figure 12. Comparison of cumulative deformation reconstructed by IDW Ordinary Kriging and the improved MLS-SVR after removing PS points in (a-c) Region I, (d-f) Region II, (g-i) Region III and (j-l) Region VI. Scatterplots of reconstructed cumulative deformation from (m) IDW, (n) Ordinary Kriging and (o) Improved MLS-SVR against PS-InSAR measurements.
![Figure 12. Comparison of cumulative deformation reconstructed by IDW Ordinary Kriging and the improved MLS-SVR after removing PS points in (a-c) Region I, (d-f) Region II, (g-i) Region III and (j-l) Region VI. Scatterplots of reconstructed cumulative deformation from (m) IDW, (n) Ordinary Kriging and (o) Improved MLS-SVR against PS-InSAR measurements.](/cms/asset/3efbcbae-80b3-467d-83cb-e56f06fa7529/tgrs_a_2230689_f0012_oc.jpg)
Figure 13. Comparison the time-series deformation between measured from PS-InSAR and reconstructed from IDW, Ordinary Kriging, and Improved MLS-SVR on point (a-d) P1-P4.
![Figure 13. Comparison the time-series deformation between measured from PS-InSAR and reconstructed from IDW, Ordinary Kriging, and Improved MLS-SVR on point (a-d) P1-P4.](/cms/asset/5e786cd9-3a30-4dc8-9751-b84956d99ac3/tgrs_a_2230689_f0013_oc.jpg)
Figure 14. Cumulative deformation of (a) PS points derived from PS-InSAR, (b) NoData grids derived from MLS-SVR, and (c) the subsidence area merged by (a) and (b); (d, g, j, m) are PS points in Region I-IV; (e, h, k, n) are NoData grids in in region I-IV; (f, i, l, o) are the merged results in Region I-IV.
![Figure 14. Cumulative deformation of (a) PS points derived from PS-InSAR, (b) NoData grids derived from MLS-SVR, and (c) the subsidence area merged by (a) and (b); (d, g, j, m) are PS points in Region I-IV; (e, h, k, n) are NoData grids in in region I-IV; (f, i, l, o) are the merged results in Region I-IV.](/cms/asset/7187f70b-133f-4b72-aefd-8ae939eaff4d/tgrs_a_2230689_f0014_oc.jpg)
Figure 15. Cumulative deformation of (a, d) PS points derived from PS-InSAR, (b,e) reconstruction final result in Area a and Area B. (c, f) Profile analysis in regions a and B.
![Figure 15. Cumulative deformation of (a, d) PS points derived from PS-InSAR, (b,e) reconstruction final result in Area a and Area B. (c, f) Profile analysis in regions a and B.](/cms/asset/cd03a2a3-dd8c-4a65-8346-dbe4214f02ae/tgrs_a_2230689_f0015_oc.jpg)
Figure 16. The change of R2 and RMSE of cumulative deformation after removing one feature type (NoC- No compressib layer; NoG-No Groundwater; NoN-No NDBI; NoF-No Fault; NoS-No subway;) or with only one feature type (OnlyC-Only compressib layer; OnlyG-Only Groundwater; OnlyN-Only NDBI; OnlyF-Only Fault; OnlyS-Only subway).
![Figure 16. The change of R2 and RMSE of cumulative deformation after removing one feature type (NoC- No compressib layer; NoG-No Groundwater; NoN-No NDBI; NoF-No Fault; NoS-No subway;) or with only one feature type (OnlyC-Only compressib layer; OnlyG-Only Groundwater; OnlyN-Only NDBI; OnlyF-Only Fault; OnlyS-Only subway).](/cms/asset/b0978604-33da-4696-ada9-3ce6b1e1078a/tgrs_a_2230689_f0016_oc.jpg)
Supplemental Material
Download MS Word (49.8 MB)Data availability statement
Upon reasonable request, the corresponding author, Professor Xiaojuan Li, Ph.D., can provide the data that support study’s conclusions.