https://orcid.org/0000-0002-1106-7505
Figures & data
Figure 1. Resilience phases of a system (Bešinović Citation2020).
Robustness: a system’s ability to resist disturbances and maintain performance. Vulnerability: performance loss due to disruption. Survivability: rate at which a system degrades. Response: the actions required to achieve a steady-state in performance. Recovery: a system’s ability to return to its original performance.
Table 1. Definitions of resilience from a data-driven perspective summarized from literature.
Figure 2. Map of London railway stations and bus stations.
A map of London railway stations and bus stops relative to the population density within London. Railway stations are in blue, and bus stops are in green. The population density in gray represents the monocentric distribution of London’s residents.
Figure 3. Trend of total weekly entries.
Red dotted line is 23rd March 2020, when restrictions were first imposed. Blue dotted lines indicate the horizon of recovery to be analyzed: from 8th June 2020 to 7th September 2020.
Table 2. Explanatory features used in this analysis.
Table 3. Summary of regression models utilized.
Figure 4. Lockdown lifted: trend of percent actual trips versus counterfactual estimate.
Trend line of actual versus estimated trips over the analysis time horizon. Individual MSOAs are in gray, and the average trend is in blue.
Figure 5. Boxplot of SRM by subregion.
MSOAs are divided by inner and outer regions. MSOAs in the Inner East and Inner West have lower spatial resilience than Outer MSOAs, although Outer West and North West are more similar to Inner MSOAs.
Figure 6. LISA.
The LISA spatial plot includes only those MSOAs that have statistically significant clustering.
Table 4. Summary of regression models.
Figure 7. Bivariate analysis of most significant features.
Plot of spatial resilience measure and top three significant features based on bivariate Moran’s I. Similar to before, each MSOA has been categorized into high–high, low–high, low–low, and high–low clusters. For example, from the total income plot, a low–high cluster indicates an area with less than average income level in an area that is above the average spatial resilience.
Figure C1. Histogram of R-Squared for all linear regression models.
This histogram provides the R-squared for every linear regression model utilized to calculate the slope of the RR.
Figure C2. Histogram of linear regression slope and Theil–Sen Slope estimator.
The full dataset versus those where the linear regression R-squared was less than 0.92. The histogram on the left compares the two slope estimators across the entire dataset. The distributions are largely similar. The histogram on the right compares the two slope estimators where the linear regression R-squared is less than 0.92. The distributions are still similar, although the Theil–Sen slope estimator has a slightly smoother distribution for slope estimates between 2 and 4.
