How computer simulation techniques can be used to measure population exposure to high levels of PM2.5 in London

by Hyesop Shin, University of Glasgow

This article is republished from AirQualityNews.com. Read the original article.

Exposure to air pollution has been extensively studied in order to understand how the health consequences of air pollution on populations can vary greatly by one’s socioeconomic profile.

Various studies have indicated that certain populations, such as the young, the elderly, socially disadvantaged individuals, and those residing near in polluted areas (e.g. industrial sites, roads), face a higher likelihood of exposure and associated health risks. In a recent report focusing on London’s air pollution, it was found that the most deprived communities were approximately 6% more susceptible to increased levels of PM_2.5 compared to the least deprived areas. Although efforts have been made to reduce annual pollutant concentrations and population exposure in London since 2013, including the implementation of the citywide Ultra Low Emission Zone (ULEZ), there are still existing disparities in exposure, particularly in more deprived areas, as well as among different age and ethnic groups. Hence, it is crucial to address these disparities comprehensively.

Epidemiological studies consistently show that exposure to particulate air pollution has harmful effects on health. To measure this exposure, the London Atmospheric Emissions Inventory (LAEI) provides a pollution map and calculates average PM_2.5 levels at people’s home addresses. However, this method has limitations because it doesn’t take into account individuals’ actual travel patterns. To overcome this limitation, researchers have used GPS technology to estimate exposure levels by combining travel surveys, GPS data, and pollution-detecting devices. This approach provides more accurate estimates by considering people’s movements throughout the day. However, there are challenges related to data quality, data cleaning, privacy, and consent, which may restrict the amount of information that can be obtained from GPS research.

To bridge the gap between atmospheric and GPS approaches, this study used a bottom-up computational method called agent-based modelling (ABM) to investigate how socioeconomic backgrounds, mobility patterns, and cumulative PM_2.5 exposure affect the population of London over time.

Spatial agent-based modelling and its application to Air Pollution Exposure

We developed a spatial ABM that creates a virtual representation of London, where individuals move throughout the city while being exposed to PM_2.5. ABMs enable us to observe the spatial and temporal variations of this pollutant in London and track people’s trajectories as pollution levels increase.

ABM is a computational method that helps us understand how individual agents, such as people or entities, interact with each other and their surroundings, creating patterns and behaviours that impact public health and society. When ABM is used in a geographic context, referred to as “Spatial ABM,” the model takes into account the physical space, allowing agents to move based on their coordinates.

Our model, called LondonHealthSim, takes the sampled population data from the 2011 Census for London boroughs and simulates the daily movement patterns. Due to the model’s capacity limitations in handling thousands of heterogeneous entities, we simplified their movements to two coordinates: home and work (or school), origin and destination matrix 2011.

Among the numerous monitoring stations located across the Greater London Area, we collected a series of hourly PM_2.5 measurements from the nearest background stations in each district. These measurements were aggregated based on home hours (assumed to be 20:00-08:00) and working hours (09:00-19:00), enabling us to capture the daily variation in air pollution exposure over an extended period.

Depending on their socioeconomic status, individuals will experience a decline in health when exposed to PM_2.5 levels above a threshold of 25µg/m₃, which is the UK’s hourly air quality standard. When the ambient pollution level exceeds this standard, the health loss function is activated. Agents exposed to PM_2.5 above the threshold will experience a non-linear decrease in health. For instance, an agent with a nominal health of 50 will lose health more rapidly than an agent with a nominal health of 100.

The baseline model was calibrated using the NHS Hospital Admitted Patient Dataset, and additional scenario experiments were subsequently conducted.

Simulating Population Exposure to PM_2.5 and the Potential Health Effects

The findings obtained from the baseline model indicate that a significant percentage (5-16%) of the population at risk in London is regularly exposed to high levels of PM2.5 during autumn and spring, with the majority falling within the age range of 15 to 64. However, the elderly population (over 65) exhibits higher vulnerability, with peaks reaching 40% experiencing PM2.5 levels surpassing the legal limit.

On average, more than 1,200 elderly individuals in the model with severe health conditions are admitted to hospitals multiple times due to air pollution. These findings highlight the existence of health disparities among specific demographic groups when exposed to prolonged and extreme levels of pollution. It underscores the importance of prioritizing assistance for the most vulnerable populations residing in the most affected areas.

The at-risk population displays significant variations across different boroughs. Boroughs such as Barking and Dagenham, Hackney, and Newham have an at-risk rate of approximately 80%. In contrast, boroughs like Bexley, Wandsworth, and Kingston upon Thames exhibit rates below 0.6%. These discrepancies can be attributed to the occurrence of multiple pollution episodes but, more importantly, to the resilience of individuals and their proximity to the traffic network.

Scenario Experiment: “What if PM_2.5 decreased by 40% but the legal limit got stricter?”

The article also experimented with a what-if scenario in which PM_2.5 was reduced by 40% but a more rigorous regulatory guideline of 10µg/m³ was established. However, even with this reduction, 6% of the population remains at risk, particularly 10% of those over 65 and 7% of those under 15. This raises concerns as these individuals continue to reside in areas that exceed the recommended World Health Organization guideline. Additionally, the at-risk rate could increase when the new regulation of 5µg/m³ takes effect in 2050. However, the findings provide evidence to support the implementation of stricter regulations in the UK, such as low-emission zones and Net Zero policy.

The paper suggests that future research could explore other behavioural guidelines for agents, such as avoiding outdoor activities during poor air quality or implementing preventive measures to mitigate health implications. It also highlights the importance of long-term planning and preparedness in utilising ABM effectively.

About the author: Hyesop Shin is a Research Associate at the MRC/CSO Social and Public Health Sciences Unit, University of Glasgow with interests in agent-based modelling (ABM), urban air quality, traffic modelling, and physical activity. His current project investigates children’s mobility patterns in Scotland using computer pathfinding algorithms, GPS, and agent-based modelling to better understand various levels of physical activity (ABM). Another aspect of his work delves deeply into the development of a traffic simulation of Glasgow in order to comprehend the displacement of congestion and air pollution as a result of the city’s low emission zone enforcement.