|Novellino, A, Terrington, R, Christodoulou, V, Smith, H and Bateson, L. 2019. Ground Motion and Stratum Thickness Comparison in Tower Hamlets, London. British Geological Survey Internal Report, OR/19/043.|
InSAR data shows that the main deformation area is localised to a restricted area, ~16 km2, between Tower Hamlets and Newham London Boroughs.
Here, previous studies (Cigna et al., 2015) have shown that changes in groundwater management in the LMBE and TAB and engineering work in the ALV and AMG have been responsible for ground motions during the ‘90s and 2000s followed by the settlement, in the range between 3 mm and 30 mm, induced by the excavation activities of the Crossrail project (Benoît, 2010; Milillo et al., 2018). The latter is the largest underground excavation plan in Europe as of 2018 and encompasses the creation of 42 km twin-bore tunnels between 2012 and 2015 stretching from west to east London.
The end of the depressurisation activities in the TAB and the Chalk Group connected to the termination of the Crossrail project boring have been proven to be the cause of ground motions observed in the Sentinel-1 data (Bonì et al., 2018).
Despite most of the AoI being stable during the analysed time span, displacement rates up to 20 mm/yr characterize the Canary Warf area where KPGR, LC, LMBE and TAB occur (Figure 7) with only AMG and TAB overlapping with all the 23 245 MPs in the area (Table 2).
|units||no of MPs|
The density plots of Figure 8 shows the distribution between the average displacement motion of MPs and the underlying geology by considering the thickness of each superficial and bedrock unit along with AMG (Figure 8). All the correlation coefficients (ρ) between the variables are quite close to 0 with LASI, RTDU and especially LC showing a negative correlation.
The analysis continued by considering the comparison between the thickness of the units and the clusters with the number of clusters chosen according to the TSS as explained in Methodology. We choose a number of clusters that account for the variance of the data, these are also clusters where the addition of a further cluster does not provide a better fit for the modelling of the data. At this point the marginal gain will drop, giving a strong curvature in the graph (Figure 9).
Calculating the second derivative will give us the curvature for each point at the graph. The second derivative results shown in Table 3, indicate that the best clusters k correspond to 3 and 5 classes. Even though the second derivative of cluster 3 is shown to be the best choice of clusters k, repeating the experiment with k=3 yields a very different clustering result. This problem is known to exist in imbalanced datasets with heavily under-sampled groups (Chawla, 2009). To be more precise, from the signals shown in Figure 10, group A represents the dominant signal in the area with 81.2% of MPs showing this pattern. Group B represents the 2.8%, group C the 1% and group D and E represent the 4.6% and 11.4% respectively. Because the majority of the groups are under-sampled, the second best choice of k=5 was selected as the number of groups. This choice is shown to represent the cluster distribution of the MPs better because the resulting groups retain their distinct features and shapes with low variation each time the algorithm runs. The results therefore can be regarded as consistent.
|k clusters||Second derivative|
In terms of temporal characteristics, groups A, C and D are within the ±5 mm/yr threshold so, despite appearing slightly different, they all show overall stability. A and C both experience uplift, followed by stability and then a period of subsidence, whereas D does not experience the uplift at the start of the time series in 2015.
MPs classified as group B display a small linear subsidence trend over the 2 years period (Figure 10b). MPs in group E display a relatively rapid uplift in 2015 followed by a period of stability (Figure 10e).
Considering the number of MPs within each group, group A is always dominant over the others groups in every unit and, overall, the number of MPs in each cluster group does not relate to the occurrence of specific units (Figure 11). However, group E represents a consistent portion in ALV with 27% of the total number of MPs for this unit (Figure 11b) compared to an average of 10.8% of MPs belonging to group E for the other layers (Figure 11).
By considering the median and the Interquartile Range (IQR) values of the normalised distributions, the percentile analysis of the boxplots of Figure 11 has been performed in order to assess the relationship between the thickness and the cluster groups within each units (i.e., the colours in Figure 12) and among the different units (i.e., the symbols in Figure 12). The lack of a dispersion of the cluster groups within the same unit, in particular for LMBE and TAB, confirm the absence of correlation between the groups and the thickness. On the other hand, there appears to be a slight negative correlation between IQR and the median values for the each cluster group across the different units, however the confidence in the correlation is low because of the large spread in the data.
The majority of MPs are classified into group A and therefore assessed to be stable with the eastern side of Tower Hamlets dominated by group E (Figure 13a). Group D is mainly associated with two structures: Old Spitalfields Market and Tobacco Dock. MPs assigned to groups B and C appear to be not connected in space and therefore isolated from other MPs belonging to that group.
Temporally and spatially, group E can be straightforwardly connected with the rising groundwater level, up to 20 m, in the deep and confined aquifer of the TAB and the Chalk Group (Figure 13b) following the end of the bulk of Crossrail dewatering activities in the area from August 2015 (Crossrail, 2016). It should be noted that, with the termination of Crossrail dewatering, the drawdown effect induced by Crossrail has now fully dissipated and that the dewatering works have complied with the obligations as enshrined in Section 46 of the Crossrail Act (2008), available at: https://www.legislation.gov.uk/ukpga/2008/18/contents.
- CIGNA, F, JORDAN, H, BATESON, L, MCCORMACK, H, and ROBERTS, C. 2015. Natural and Anthropogenic Geohazards in Greater London Observed from Geological and ERS-1/2 and ENVISAT Persistent Scatterers Ground Motion Data: Results from the EC FP7-SPACE PanGeo Project. Pure Appl. Geophys. 172, 2965–2995. https://doi.org/10.1007/s00024-014-0927-3
- BENOIT, J. 2010. Low-volume-loss tunnelling for London ring main extension. Proceedings of the ICE — Geotechnical Engineering, 163 (3). doi:10.1680/geng.2010.163.3.167
- MILILLO, P, GIARDINA, G, DEJONG, M J, PERISSIN, D, and MILILLO, G. 2018. Multi-Temporal InSAR Structural Damage Assessment: The London Crossrail Case Study. Remote Sensing, 10(2). 287. https://doi.org/10.3390/rs10020287
- BONI, R, BOSINO, A, MEISINA, C, NOVELLINO, A, BATESON, L, and MCCORMACK, H. 2018. A Methodology to Detect and Characterize Uplift Phenomena in Urban Areas Using Sentinel-1 Data. Remote Sensing, 10(607). https://doi.org/10.3390/rs10040607
- CHAWLA, N V. 2009. Data mining for imbalanced datasets: An overview. In Data mining and knowledge discovery handbook. Springer, Boston, MA, 875–886.
- CROSSRAIL, 2016. Crossrail Project Dewatering Works—Close-out Report. Available online: https://learninglegacy.crossrail.co.uk/wp-content/uploads/2017/01/7A-026_1-Dewatering-Close-out-report.pdf (accessed on 12 November 2017).