This paper explores the carbon impacts of the travel restrictions placed on the general population in the UK as a result of Covid-19 pandemic. Specifically, it focusses on commute journeys where the advice from the Government was to work from home, if possible (GOV.UK 2020). The article uses anonymised and aggregated data from O2, a major mobile network provider, between Feb-2020 to Jun-2020 to understand how commute trips fluctuated. This is of relevance to the broader climate crisis which requires deep cuts in emissions from significant demand reduction measures over the next two decades (Brand, Anable, and Morton 2019). Our research question is:
What impact did the COVID-19 “work from home if you can” policy have on the commute patterns and CO2 emissions?
Anonymised and aggregated mobile data insights for Feb-2020 to Jun-2020, from O2 mobile, was used to evaluate the spatial variations in commute journeys. O2 connects with 23 million devices in the UK (25% market-share) offering movement insights, when connecting with the mobile network. Daily estimates of numbers of commute journeys were provided for each of 375 Local Authority Districts (LADs) based on analysis of the durations of phone locations away from home which matched a commute pattern. Analysis of monthly commute data from the National Travel Survey [NTS0504] shows that the number of commute trips per day in Feb-2020 was similar to other months in the dataset, prior to the lockdown (Office of National Statistics 2020b). Therefore, the data for February was treated as the “pre-lockdown” baseline. Lockdown began on 23rd March in the UK and so April to June were considered “during-lockdown”.
To convert journey reductions into a CO2eq, it was necessary to assign estimates of journey length (by region) and mode split to each local authority. The NTS data, however, allows for differentiation at a regional level and only for pre-lockdown scenario. Similar data was sourced from the Scottish Transport Statistics (Transport Scotland Analytical Services 2018), Welsh Travel statistics (Welsh Government Statistics and Research 2019) and the Northern Ireland Transport Statistics (Northern Ireland Statistics and Research Agency 2020) for UK-wide coverage. Regional figures, assigned to the local authorities, within that region, are shown in Table 1.
There are a few shortcomings to the data and assumptions we have adopted. It is neither possible to estimate the actual monthly mode share reductions for commute, over the lockdown period, nor is it, to determine the extent to which the public transport (PT) levels reduced, by month and locality. We have attempted to allow for the impacts of these uncertainties through a scenario analysis described further below. Other factors such as increased unemployment or people being furloughed were excluded and so the estimates must be seen as upper bounds of CO2 reduction.
The district-level mobile data was first normalised to each of the 375 districts using the population estimates from UK census (Office for National Statistics, National Records of Scotland, and Northern Ireland Statistics and Research Agency 2017). The districts were then grouped into the NTS-geographic regions (shown in Table 1) and the commute journeys were differentiated into the NTS-modal shares. Upon splitting the trips, two other parameters were factored into commute CO2 estimation (Figure 1): the average trip lengths by car/ other modes and modal emission factors.
Monthly commute CO2 emissions were estimated for each of the local authorities across the UK, between Feb-2020 and Jun-2020. February emissions were set as reference and the monthly percentage changes in CO2 emissions, between Apr-2020 and Jun-2020, were estimated for “during lockdown”.
To deal with the uncertainties in the estimated emissions, a 2 x 2 matrix was created which allows for variations in the commute mode shares and the amount of public transport that may have operated, before and after the lockdown. For clarity, the two mode share assumptions were: Scenario-a: “Pre-lockdown” mode share pattern retained; Scenario-b: “During-lockdown” estimated mode share patterns (details in “Notes and Assumptions” in Table 2). Considering the potential for emissions from PT, it is assumed that the emissions from commuters who usually travelled by PT (in “pre-lockdown” settings) would be saved “during-lockdown” as the services reduced (Scenario-c). Alternatively, Scenario-d included emissions assuming that services may still have operated “during-lockdown”.
The anticipated baseline emissions determined for February 2020 (before lockdown) within this study, was 1422kT CO2. A validation of our baseline monthly commute CO2 emissions estimate for February, against another national estimate for monthly commute emissions, is within 5%. Overall, the range of emissions, reported in this sensitivity analysis, correspond to just 30-38% of an equivalent 3-monthly pre-lockdown baseline emissions (1422kT estimated from February figures). Relative to overall magnitude of the reduction, the impact of mode share variations between the scenarios are relatively small. Assuming the mode share to be more car-oriented in lockdown, in line with national average mode use, estimates results in emissions which are 47.8kT to 62.3kT higher over the three months (or 3.4% - 4.4% of the anticipated baseline). Whether or not reductions in public transport translated into savings in public transport emissions, made a difference of 57.8-72kT or 4-5% of the anticipated baseline.
Figure 2 represents the monthly fluctuations in district-level emissions of commute carbon (representing scenario pair a-d), for Mar-2020 to Jun-2020. Commute traffic increased in the starting weeks of March than normal (20-25%) which is most likely to be in preparation for the looming lockdown. A sharp drop then occurred from precautionary local closures of businesses, in advance of the 23rd March lockdown. However, the highest levels of commute carbon reduction, were observed in May 2020, in the range of -40 to -78%. Applying a scaling factor based on the number of workdays to account for potential commute in a standard month, across March to June, led our calculated monthly commute emissions to be ±5% that of our baseline estimates (1500kTCO2eq /month).
A matching of the travel adaptation and carbon reduction data was then made with district-level data on the density of certain industrial sectors that could successfully adapt to operate virtually and their associated car-ownership patterns. For this assessment, the 375 LADs were grouped into quintiles based on the percentage of commute emission reductions they demonstrated (Table 3). The openly available MOT (annual vehicle test) dataset was used to estimate the proportion of the car fleet held by households in each quintile (Department for Transport 2020). As for employment type, the UK labour statistics demonstrated that of the 30-45% employed in professional, associate professional/ technical occupation, roughly 91.1% and 86.1% worked from home during the lockdown (Office of National Statistics 2020a). Sectors that followed this pattern also include information/ communication and scientific/ technical activities (of whom 79% and 45% worked remotely). Table 3 shows how the percentages of industry type compare across the quintiles.
The top 20% districts for CO2 emission reduction (avg. of 68% to 78%) were those with the highest proportion of specific industry roles which could be worked from home and held the highest proportion of the UK fleet, thus reinforcing the effects of commute carbon reduction over the lockdown period (Table 3).
Future work should address the extent to which forced behavioural shifts can become embedded in the commute demand reduction and wider decarbonisation policies. There will be a need to explore what this means for different places, once a more nuanced understanding of the post-pandemic mode-shift behaviours begin to take shape.
This project has received funding from EPSRC under grant agreement EP/S032002/1 and EP/R035288/1. This project has also received data courtesy of O2. The article is the responsibility of the authors and does not imply endorsement by the funders.
The trip data used in this study covers changes implied by only the N of 23 million devices that connect to the O2 network and not the entire population
Commutes are estimated from regular home and non-home pairings with extended durations. A minimum of 10 people moving between an OD-pair is required to classify as a commute journey, to protect the anonymity of users.
A majority of the LADs (roughly 60% of the 375 LADs) have shown a relative spike in commute trips in the first two weeks of the March 2020 (supplemented by higher no. of working days), compared to their daily estimated equivalents in February 2020 (baseline month) (please see figure 2) .
These estimates were drawn from averaging the daily emissions relative to their equivalent in Feb-2020. our CO2 estimates could be validated through a comparison with publicly reported figures for commute emission. It is estimated that the UK’s annual commute carbon amounts to 18 billion kgCO2eq, which breaks down to roughly 1500 kTCO2eq per month (Mobilityways 2021)
MOT dataset provides district-level information including the total numbers of vehicles by make/model and fuel type.