Modelling the carbon impacts of EV uptake

How we used Agent-Based Modelling to analyse decarbonisation policies

This is the second post in our series on Transport East’s (TE) BERTIE Agent-Based Model (ABM). In this post, we’ll take a closer look at the findings of the electric vehicle (EV) scenarios we modelled.

Table of Contents
Context
Key Findings
Diving Deeper
Show me the outputs!
Closing Thoughts

Context

Transport East was created in 2018, serving as the sub-national transport body for Norfolk, Suffolk, Essex, Southend-on-Sea, and Thurrock. There are 3.5 million residents in the TE region; potentially increasing to 4 million residents by 2041. The transport sector contributes to 42% of carbon emissions in the region; primarily from road transport. In July 2022, Transport East published their Transport Strategy and identified the below four goals to meet decarbonisation targets by 2040.

TE’s decarbonisation goals to achieve Net Zero by 2040

TE is ready to deliver on these goals, recognising the value of a disaggregated, multi-modal approach to model and evaluate decarbonisation pathways.

As our previous blog discussed, we developed requirements for ABM scenarios that include increased driving costs, EV uptake, active scenarios, and a combination of these three. Our baseline assumes 33% uptake, and the EV scenarios modelled are EV Double, which assumes 66% uptake and EV High at 88%. Both scenarios assumed a disparity in uptake across income levels.

All scenarios were modelled in isolation (with the exception of the “combined” scenario) to distinctly demonstrate the behavioural responses. The designed scenarios closely align with the aims of TE’s Goals 2 and 3, Shift Modes and Switch Fuels. Our modelling found that scenarios with increased EV uptake will yield some of the greatest reductions in carbon emissions (29%–58.9%).

Emissions reductions relative to the baseline. Includes both household and freight emissions.

Key Findings

Despite the EV scenarios providing the clearest path to decarbonisation, we found a few unintended knock-on effects in the behavioural response. In scenarios with higher EV uptake, vehicle km increased relative to the baseline (an estimated increase ranging from 0.8% to 1.2%).

Vehicle km is a product of the number of cars on the road and the distance they travel. An increase indicates either more trips, longer trips, or both. Despite the small change, we think this is important as all other scenarios (active, increased driving costs) decreased vehicle km. Furthermore, we found people in our model chose to walk less frequently and drive more. In the EV High scenario, walking trips decreased by 1% whereas car trips increased by 0.6%.

Less surprisingly, we found that residents from lower-income households have a relatively lower capability to decarbonise — assuming the cost of transitioning to EV is solely undertaken by the household. Research demonstrates that higher-income households within the UK ([1], [2]) and other international cities ([3]) tend to own a larger share of electric vehicles.

If EVs are the only decarbonisation solution, there is a risk that only certain households will be able to decarbonise, and fewer trips will be made by active modes. Furthermore, HGVs do not have a decarbonisation pathway in our models, increasing their share of the resulting carbon emissions by 2040.

These findings validate TE’s desire to have a broad range of decarbonisation approaches in their region — pursuing only EVs is unlikely to provide a holistic solution. The combination of several policies applied in the right contexts might get TE closer to a more equitable decarbonised future.

Diving Deeper

So how exactly did we arrive at these findings? In ABMs, we model individuals and their choices. This enables us to tease out behavioural changes and policy impacts across different demographics and explore a range of questions. To do this responsibly, we made a couple of assumptions, outlined below:

— Utilised UK Transport Appraisal Guidance (TAG) for EV uptake and emissions. The May 2022 publication estimates 33% of households will have EVs by 2040, which became our baseline model input. TAG assumes the emissions factors for EVs are based on generation emissions and modelled as proportional to distance travelled; independent of speed (unlike fuel vehicles). We used Pandia to estimate emissions and generate the below curves demonstrating CO2e at different speeds across vehicle types.

[1] Comparison of emissions curves for all vehicle types in the model, assuming a distance travelled of 1 km. Bus emissions are highest, whereas EV vehicle types have the lowest emissions. [2] A zoomed in chart for the EV vehicle types to demonstrate the constant relationship modelled in the ABM.

— Built an assumption of EV uptake across socioeconomic groups. We want to help our clients find equitable solutions and therefore chose to model the disparity in EV uptake. To do so, we consulted the National Travel Survey (NTS) of household vehicle ownership to model a marginal disparity amongst households in the uptake of EVs by 2040.

Show me the outputs!

When delivering work for our clients, we aim to provide outputs from a variety of perspectives: carbon emissions, mode shift, and spatial analysis. Below is a small sample of what we showed TE.

Looking at household emissions, we see that lower-income households contribute the least (25%) to total emissions in the baseline; due to their utilisation of active and sustainable modes (public transport, cycling, and walking) to complete daily activities.

We also see high-income households contribute fewer emissions than medium-income households due to the assumption they will have a relatively higher uptake in EVs. In the EV uptake scenarios (66%, 88% uptake), we find this assumption gives high-income households the greatest potential to decarbonise relative to the baseline.

Emissions percent share for each income household (“hhincome”) by scenario, as EV uptake increases, emissions share for high income households decreases significantly.

We concluded that without additional support, lower-income households will only be able to reduce their emissions per capita by 63% whereas high-income households will be able to reduce by 90%. Our modelling helps demonstrate the potential magnitude of inequity in a household’s ability to decarbonise by transitioning to EVs, highlighting a need for corrective policies.

We also found that people switched from active modes to car. This outcome reflects the incentives behind behaviour change — an electric vehicle has relatively lower running costs than fuel, incentivising an increase in driving to reduce travel time. Below we have plotted Sankey diagrams of the shifts in mode: the full set of mode shifts on the left and net mode shifts on the right.

When looking at the net shift, we can see that overall the majority of shifted trips choose car, and there is very little movement away from the car mode. Short trips (0–5km) may have more potential to mode shift away from cars, and we found that both EV scenarios were the only ones to decrease active travel and public transport modes for short distances.

[1] Absolute mode shift from baseline trips to EV High trips. We see that that the majority of trips that went as car in the baseline stay as car in the EV High scenario. [2] Net mode shift of trips. We can clearly see the majority of trips that shift towards car come from bus, rail, and walking trips.

In the EV scenarios, we see overall emissions decrease; for HGVs, however, emissions stay roughly flat. We instead find spatial differences, and can examine the links where emissions increase or decrease for HGVs. This gives TE an idea of which corridors are likely to have increased emissions either due to congestion on those links or re-routing behaviour to avoid said congestion.

With this, TE could target the specific areas where they see an undesirable change in behaviour and devise low-carbon alternatives for HGVs (such as incentivising a shift to rail). This analysis highlights the importance of simulating the interactions between different types of road users, and how our BERTIE build enables us to interrogate simulations to a fine level of detail.

Difference in Carbon Emissions for HGVs: normalised by length of each link. Green indicates where carbon emissions reduced, red where emissions increased.

Closing Thoughts

The EV scenarios produced a series of interesting results that highlights the complexities of decarbonisation. Switching to EVs reduces carbon emissions the most. However, it risks inequitable decarbonisation outcomes and a potential response that increases driving — conflicting with TE’s decarbonisation goal of a mode shift away from cars.

Even if it is possible to shift 88% of vehicles to EVs, it will require significant investment in charging infrastructure, supply chains, and an energy grid that can support this demand (coincidentally, we’ve developed a tool to estimate charging demand, more soon…).

Another important consideration for TE is the impact on urban vs rural communities. Research has shown rural and suburban communities in Norway own the larger share of EVs as compared to urban centres. This difference is likely due to travel needs, indicating an opportunity to design EV policies that consider these patterns.

The Transport East region has an ambitious decarbonisation strategy that proposes a variety of solutions to decarbonise their transportation network, demonstrating a systems approach of pursuing net zero. We look forward to continue partnering with TE on applications of BERTIE to explore targeted policy scenarios towards a decarbonised equitable future.

See also: Introducing BERTIE — an Agent Based Model for Transport East