Estimates of the Carbon Impacts of Commute Travel Restrictions due to COVID-19 in the UK

Findings This paper explores the carbon impacts of the reductions in commute travel which resulted from restrictions placed on the general population in the UK. The article uses anonymised and aggregated mobile data for the period February 2020 to June 2020 to understand how commute trips changed spatially. This has been linked to journey length and emissions data to produce estimates of the consequent reductions in CO 2 (an average range of 17-60%). At a local level, the key factors that contributed to substantial CO 2 reductions were high car ownership, paired with the prevalence of specific industrial employment types that could readily transition from a desk-based work to virtual working.


Methods
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) 1 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. 2 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 23 rd March 3 in the UK and so April to June were considered "during-lockdown".
To convert journey reductions into a CO 2 eq, 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 CO 2 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 CO 2 estimation ( Figure 1): the average trip lengths by car/ other modes and modal emission factors.
Monthly commute CO 2 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 CO 2 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 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) . 3 Estimates of the Carbon Impacts of Commute Travel Restrictions due to COVID-19 in the UK  Figure 1: Approach for the estimation of CO 2 emissions by mode for commute trips lockdown. For clarity, the two mode share assumptions were: Scenario-a: "Prelockdown" 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 (Scenarioc). Alternatively, Scenario-d included emissions assuming that services may still have operated "during-lockdown".

Findings
The anticipated baseline emissions determined for February 2020 (before lockdown) within this study, was 1422kT CO 2 . A validation of our baseline monthly commute CO 2 emissions estimate for February, against another national estimate for monthly commute emissions, is within 5%. 4 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 These estimates were drawn from averaging the daily emissions relative to their equivalent in Feb-2020. our CO 2 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 kgCO 2 eq, which breaks down to roughly 1500 kTCO 2 eq per month (Mobilityways 2021) Table 2: Estimation of monthly CO 2 emissions applying "pre-lockdown" and "during-lockdown" modal distribution, followed by a sensitivity study encompassing the impacts of modal shares variations and fluctuations in public transport services and usage Parameters Parameters Scenario-a: Scenario-a: "Pre-lockdown" modeshare settings 1 Notes and Assumptions for the scenarios 1 Mode shares were assumed to be in line with the pre-lockdown modal split of 68% for cars; 7% for buses; 10% for rail and 15% for other modes (Please see Table   1) (motorised travel make up 85% of total commute mode distribution; at 100%, the split will be 80% for cars, 8.23% for buses and 11.8% for rail;) 2 "During-lockdown" commute trips were acquired by applying DfTs (March-June) average mode reductions (57% for cars; 80% for buses and 90% for rail); motorised commute modes -Car: from 80% to 92.4%; Bus: from 8.23% to 4.4%; rail: from 11.8% to 3.1%) 3 March commute estimates (383kT and 426 kTCO 2 eq) were excluded from the 3-month total estimates since the travel restrictions had not perfectly set in until the last week of the month (23 rd March) leading for these CO 2 estimates to be misleading for our assumed scenarios. 4 Emissions in scenario (a-d) represents 3-month total CO 2 emissions for all modes, in the "pre-lockdown" setting, taking into account the baseline PT emissions (Feb), allowing for the assumption that all PT services were operational, despite the fall in PT commute. 5 Emissions in scenario (b-d) represents 3-month total CO 2 emissions for all modes, from the "during lockdown" commute setting, taking into account the baseline PT emissions (Feb), following PT operations mentioned in point 4. 6 Working days and School/National Holidays in 2020 by month: February (20, 5) March (23, 3), April (22, 7), May (21, 10), June (22,4) 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. potential commute in a standard month, across March to June, led our calculated monthly commute emissions to be ±5% that of our baseline estimates (1500kTCO 2 eq /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 5 was used to estimate the proportion of the car fleet held by households in each quintile MOT dataset provides district-level information including the total numbers of vehicles by make/model and fuel type.

5
Estimates of the Carbon Impacts of Commute Travel Restrictions due to COVID-19 in the UK (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.

Findings
The top 20% districts for CO 2 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.