Introducing Pandia, a Tool to Estimate Carbon Emissions
Using Agent-Based Models to drive decarbonisation
Divya is a data scientist at Arup’s City Modelling Lab, she is exploring the potential of transport decarbonisation through simulations.
Theodore is a senior data scientist at Arup’s City Modelling Lab, with a background in transport planning and economic appraisal.
This article introduces Pandia, a python tool to estimate carbon emissions utilising City Modelling Lab’s simulation outputs. We introduce the tool, why we built it, and how it enhances our platform for solution discovery.
Carbon Emissions of Road Transport
Transport emissions have remained stubbornly high over the past decades despite significant and widespread technological changes. While total emissions in the UK have decreased by 44% relative to 1990 levels (largely due to a transition away from coal power), emissions from transport only fell by 5% during the same period (DfT, 2021). To meet the ambition of net zero by 2050, policy makers are challenged with devising timely, efficient, and effective solutions in highly complex and intertwined systems of systems.
At the City Modelling Lab (CML), we are working to provide data driven insights for these decision makers to trial ideas and observe their potential impacts (both intended and unintended alike).
CML Approach
We leverage the Agent Based Modelling paradigm to simulate emerging behavioural responses to transport interventions across large spatial areas and multiple modes. By operating at a low spatial level and high temporal resolution, we simulate the individual, their choices, throughout every second in a day. This provides a foundation for Pandia, our carbon emissions tool, to interrogate who is emitting, where, and when these emissions are occurring.
Pandia is a layer of analysis built on top of our simulation outputs. It estimates carbon emissions at the individual agent level, based on the speed, vehicle type, and fuel type. Pandia provides insight into emissions behaviours and supports scenario analysis of various interventions. Our approach is to build simulations that are bottom-up, informed by existing research, and demonstrate complexity. The creation of Pandia follows these principles:
- Bottom-up: We collect data with respect to how, where, and when different individuals travel within the day to create a synthetic representation of the population. These activities are modelled through our population synthesis tool, PAM and simulated using MATSim.
- Informed by existing research: To realistically represent vehicles in the network, our input datasets include microdata from the UK’s National Travel Survey and aggregate Office for National Statistics data. To estimate emissions, we built a calculation methodology based on research and appraisal guidance (link) that enables us to emulate real-world processes.
- Demonstrates complexity Pandia enables us to capture speed and emissions for a single agent throughout their entire journey for a simulated day. The next four plots generated with Pandia demonstrate the range of insights for a simulated agent, Agent 24, in London.
Agent 24’s Journey in London
Agent 24 takes two buses to get from the gym to a medical centre, with different levels of occupancy in each bus. We are able to account for the number of people sharing the bus with Agent 24 to demonstrate their personal emissions throughout the journey.
The first plot of bus occupancy illustrates the number of people on the bus with Agent 24; the first bus carries 10 people whereas the second bus carries 20. The second plot shows cumulative emissions for Agent 24, where Agent 24’s emissions on the first bus has a steeper slope relative to the slope of cumulative emissions while riding the second bus. This change in slope demonstrates that the first bus has less people, and therefore the emissions for Agent 24 rises faster than when they are on the second bus. Finally, the third plot demonstrates the non-linear relationship between emissions and speed. Low speeds result in higher emissions per person kilometre whereas higher speeds offer diminishing emissions reductions. We can see the second bus has a lower emission per person kilometre curve; demonstrating the emissions efficiencies gained with fuller buses.
The map below shows the associated emissions along each road of their journey; the width illustrates the higher emissions on the first bus (red) relative to the second bus (blue). In this example, the outputted calibrated model showed the agent choosing to take two different buses from Nine Elms Lane to the A3036 and cross over Putney Bridge.
This ability to demonstrate complex emissions behaviour allows us to understand when and where agents are travelling in the most energy efficient mode. It forms our basis to explore the chains of daily choices of agents and their response to proposed interventions to achieve Net Zero.
High Fidelity
Although we originally attempted estimating emissions at an aggregate level, we found that estimating emissions at a higher fidelity captures a holistic view of transport and travel behaviour. Our high fidelity method allows us to disaggregate emissions from multiple perspectives: equity, temporal, and spatial. This enables modelling of relationships between emissions and household attributes, travel patterns, or types of vehicles. Pandia can support a range of nuanced solution discovery. Below are examples of questions we aim to explore from a carbon emissions perspective.
This multi-perspective view of carbon emissions sets it apart from existing aggregated methods by precisely identifying targeted areas to reduce emissions. And yet, to imperfectly quote Uncle Ben: ”With great fidelity comes great responsibility”. We are constantly straddling the balance of precision versus accuracy in our modelling. This precision allows us to reframe how we think about managing transport carbon emissions. No longer do we have to build solutions for a hypothetical average person, we can now explore impacts on each agent within our system. We can observe how complex individual interactions emerge to unexpected results. This increased fidelity therefore requires taking care to generate outputs that are both insightful and representative.
It is the beginning of an exciting journey to discover how our modelling may support decisions that yield a reduction in transport emissions. The next stage of this journey requires standardisation of our data outputs and visualisations. We will also generate scenarios of future years to provide insight into behaviour and fleet changes that may help achieve Net Zero by 2050. Finally, we are investigating the potential of developing Pandia into a standalone solution separate from our simulations.
If you would like to make use of Pandia or have more questions then please get in touch — citymodelling@arup.com
