Using Explainable Machine Learning to Interpret the Effects of Policies on Air Pollution: COVID-19 Lockdown in London.

Activity changes during the COVID-19 lockdown present an opportunity to understand the effects that prospective emission control and air quality management policies might have on reducing air pollution. Using a regression discontinuity design for causal analysis, we show that the first UK national lockdown led to unprecedented decreases in road traffic, by up to 65%, yet incommensurate and heterogeneous responses in air pollution in London. At different locations, changes in air pollution attributable to the lockdown ranged from -50% to 0% for nitrogen dioxide (NO2), 0% to +4% for ozone (O3), and -5% to +0% for particulate matter with an aerodynamic diameter less than 10 µm (PM10), and there was no response for PM2.5. Using explainable machine learning to interpret the outputs of a predictive model, we show that the degree to which NO2 pollution was reduced in an area was correlated with spatial features (including road freight traffic and proximity to a major airport and the city center), and that existing inequalities in air pollution exposure were exacerbated: pollution reductions were greater in places with more affluent residents and better access to public transport services.

Targeted Use of Sustainable Aviation Fuel to Maximize Climate Benefits.

Sustainable aviation fuel (SAF) can reduce aviation's CO2 and non-CO2 impacts. We quantify the change in contrail properties and climate forcing in the North Atlantic resulting from different blending ratios of SAF and demonstrate that intelligently allocating the limited SAF supply could multiply its overall climate benefit by factors of 9-15. A fleetwide adoption of 100% SAF increases contrail occurrence (+5%), but lower nonvolatile particle emissions (-52%) reduce the annual mean contrail net radiative forcing (-44%), adding to climate gains from reduced life cycle CO2 emissions. However, in the short term, SAF supply will be constrained. SAF blended at a 1% ratio and uniformly distributed to all transatlantic flights would reduce both the annual contrail energy forcing (EFcontrail) and the total energy forcing (EFtotal, contrails + change in CO2 life cycle emissions) by ∼0.6%. Instead, targeting the same quantity of SAF at a 50% blend ratio to ∼2% of flights responsible for the most highly warming contrails reduces EFcontrail and EFtotal by ∼10 and ∼6%, respectively. Acknowledging forecasting uncertainties, SAF blended at lower ratios (10%) and distributed to more flights (∼9%) still reduces EFcontrail (∼5%) and EFtotal (∼3%). Both strategies deploy SAF on flights with engine particle emissions exceeding 1012 m-1, at night-time, and in winter.

Dynamics and outcomes of binary collisions of equi-diameter picolitre droplets with identical viscosities.

The dynamics of binary collisions of equi-diameter picolitre droplets with identical viscosities, varying impact speeds and impact angles have been investigated experimentally and compared to collision outcome prediction models. Collisions between pairs of pure water droplets with a viscosity of 0.89 mPa s and pairs of aqueous-sucrose (40% w/w) droplets with a viscosity of 5.17 mPa s were examined. The colliding droplets were ∼38 µm in diameter, which is around ten times smaller than those previously investigated when examining the effect of viscosity on the outcome of binary droplet collisions. Varying the impact speed and angle resulted in different collision outcomes, including coalescence, reflexive separation and stretching separation. The collision outcomes were plotted on two viscosity dependent regime maps. The regime boundaries are generally in agreement with earlier literature for both high and low viscosity cases. The agreement between experiment and theory, for both fluids, gives more confidence in the models tested here to predict collision outcomes for droplets of this size and these viscosities.

Source terms for benchmarking models of SARS-CoV-2 transmission via aerosols and droplets.

There is ongoing and rapid advancement in approaches to modelling the fate of exhaled particles in different environments relevant to disease transmission. It is important that models are verified by comparison with each other using a common set of input parameters to ensure that model differences can be interpreted in terms of model physics rather than unspecified differences in model input parameters. In this paper, we define parameters necessary for such benchmarking of models of airborne particles exhaled by humans and transported in the environment during breathing and speaking.

Long-Term In-Use NOx Emissions from London Buses with Retrofitted NOx Aftertreatment.

Buses constitute a significant source of air pollutant emissions in cities. In this study, we present real-world NOx emissions from 97 diesel-hybrid buses measured using on-board diagnostic systems over 44 months and 6.35 million km in London. Each bus had previously been retrofitted with a selective catalytic reduction (SCR) aftertreatment system to reduce emissions of nitrogen oxides (NOx). On average, parallel hybrid (PH) and series hybrid (SH) buses emitted 3.80 g of NOx/km [standard deviation (SD) of 1.02] and 2.37 g of NOx/km (SD of 0.51), respectively. The SCR systems reduced engine-out emissions by 79.8% (SD of 5.0) and 87.2% (SD of 2.9) for the PHs and SHs, respectively. Lower ambient temperatures (0-10 °C) increased NOx emissions of the PHs by 24.2% but decreased NOx emissions of the SHs by 27.9% compared to values found at more moderate temperatures (10-20 °C). To improve emissions inventories, we provide new distance-based NOx emissions factors for different ranges of ambient temperature. During the COVID-19 pandemic, the emissions benefits of reduced congestion were largely offset by more frequent route layovers leading to lower SCR temperatures and effectiveness. This study shows that continuous in-service measurements enable quantification of real-world vehicle emissions over a wide range of operations that complements conventional testing approaches.

A novel multi-pollutant space-time learning network for air pollution inference.

Detailed information about air pollution in space and time is essential to manage risks to public health. In this paper we propose a multi-pollutant space-time learning network (Multi-AP learning network), which estimates pixel-wise (grid-level) concentrations of multiple air pollutant species based on fixed-station measurements and multi-source urban features, including land use information, traffic data, and meteorological conditions. We infer concentrations of multiple pollutants within one integrated learning network, which is applied to and evaluated on a case study in Chengdu (4900 km2, 26 April - 12 June 2019), where air pollutant (PM2.5, PM10 and O3) measurements from 40 monitoring sites are used to train the network to estimate pollutant concentrations in 4900 grid-cells (1 km2). The Multi-AP learning network allows us to estimate highly-resolved (1 km × 1 km, hourly) air pollution maps based on pollutant measurements which cover less than 1% of the grid-cells with better accuracy compared to other approaches, and with significant computational efficiency improvements. The time-cost is 1/3 of the time-cost of modelling each pollutant individually. Furthermore, we evaluate the relative importance of features and find that the meteorological feature set is the most important followed the land use features. The proposed Multi-AP method could be used to estimate air pollution exposure across a city using a limited set of air pollution monitoring sites.

The ventilation of buildings and other mitigating measures for COVID-19: a focus on wintertime.

The year 2020 has seen the emergence of a global pandemic as a result of the disease COVID-19. This report reviews knowledge of the transmission of COVID-19 indoors, examines the evidence for mitigating measures, and considers the implications for wintertime with a focus on ventilation.

Modelling of instantaneous emissions from diesel vehicles with a special focus on NOx: Insights from machine learning techniques.

Accurate instantaneous vehicle emissions models are vital for evaluating the impacts of road transport on air pollution at high temporal and spatial resolution. In this study, we apply machine learning techniques to a dataset of 70 diesel vehicles tested in real-world driving conditions to: (i) cluster vehicles with similar emissions performance, and (ii) model instantaneous emissions. The application of dynamic time warping and clustering analysis by NOx emissions resulted in 17 clusters capturing 88% of trips in the dataset. We show that clustering effectively groups vehicles with similar emissions profiles, however no significant correlation between emissions and vehicle characteristics (i.e. engine size, vehicle weight) were found. For each cluster, we evaluate three instantaneous emissions models: a look-up table (LT) approach, a non-linear regression (NLR) model and a neural network multi-layer perceptron (MLP) model. The NLR model provides accurate instantaneous NOx predictions, on par with the MLP: relative errors in prediction of emission factors are below 20% for both models, average fractional biases are -0.01 (s.d. 0.02) and -0.0003 (s.d. 0.04), and average normalised mean squared errors are 0.25 (s.d. 0.14) and 0.29 (s.d. 0.16), for the NLR and MLP models respectively. However, neural networks are better able to deal with vehicles not belonging to a specific cluster. The new models that we present rely on simple inputs of vehicle speed and acceleration, which could be extracted from existing sources including traffic cameras and vehicle tracking devices, and can therefore be deployed immediately to enable fast and accurate prediction of vehicle NOx emissions. The speed and the ease of use of these new models make them an ideal operational tool for policy makers aiming to build emission inventories or evaluate emissions mitigation strategies.

Mitigating the Climate Forcing of Aircraft Contrails by Small-Scale Diversions and Technology Adoption.

The climate forcing of contrails and induced-cirrus cloudiness is thought to be comparable to the cumulative impacts of aviation CO2 emissions. This paper estimates the impact of aviation contrails on climate forcing for flight track data in Japanese airspace and propagates uncertainties arising from meteorology and aircraft black carbon (BC) particle number emissions. Uncertainties in the contrail age, coverage, optical properties, radiative forcing, and energy forcing (EF) from individual flights can be 2 orders of magnitude larger than the fleet-average values. Only 2.2% [2.0, 2.5%] of flights contribute to 80% of the contrail EF in this region. A small-scale strategy of selectively diverting 1.7% of the fleet could reduce the contrail EF by up to 59.3% [52.4, 65.6%], with only a 0.014% [0.010, 0.017%] increase in total fuel consumption and CO2 emissions. A low-risk strategy of diverting flights only if there is no fuel penalty, thereby avoiding additional long-lived CO2 emissions, would reduce contrail EF by 20.0% [17.4, 23.0%]. In the longer term, widespread use of new engine combustor technology, which reduces BC particle emissions, could achieve a 68.8% [45.2, 82.1%] reduction in the contrail EF. A combination of both interventions could reduce the contrail EF by 91.8% [88.6, 95.8%].

Real world CO2 and NOx emissions from 149 Euro 5 and 6 diesel, gasoline and hybrid passenger cars.

In this study CO2 and NOx emissions from 149 Euro 5 and 6 diesel, gasoline and hybrid passenger cars were compared using a Portable Emissions Measurement System (PEMS). The models sampled accounted for 56% of all passenger cars sold in Europe in 2016. We found gasoline vehicles had CO2 emissions 13-66% higher than diesel. During urban driving, the average CO2 emission factor was 210.5 (sd. 47) gkm-1 for gasoline and 170.2 (sd. 34) gkm-1 for diesel. Half the gasoline vehicles tested were Gasoline Direct Injection (GDI). Euro 6 GDI engines <1.4ℓ delivered ~17% CO2 reduction compared to Port Fuel Injection (PFI). Gasoline vehicles delivered an 86-96% reduction in NOx emissions compared to diesel cars. The average urban NOx emission from Euro 6 diesel vehicles 0.44 (sd. 0.44) gkm-1 was 11 times higher than for gasoline 0.04 (sd. 0.04) gkm-1. We also analysed two gasoline-electric hybrids which out-performed both gasoline and diesel for NOx and CO2. We conclude action is required to mitigate the public health risk created by excessive NOx emissions from modern diesel vehicles. Replacing diesel with gasoline would incur a substantial CO2 penalty, however greater uptake of hybrid vehicles would likely reduce both CO2 and NOx emissions. Discrimination of vehicles on the basis of Euro standard is arbitrary and incentives should promote vehicles with the lowest real-world emissions of both NOx and CO2.

Greenhouse Gas and Noxious Emissions from Dual Fuel Diesel and Natural Gas Heavy Goods Vehicles.

Dual fuel diesel and natural gas heavy goods vehicles (HGVs) operate on a combination of the two fuels simultaneously. By substituting diesel for natural gas, vehicle operators can benefit from reduced fuel costs and as natural gas has a lower CO2 intensity compared to diesel, dual fuel HGVs have the potential to reduce greenhouse gas (GHG) emissions from the freight sector. In this study, energy consumption, greenhouse gas and noxious emissions for five after-market dual fuel configurations of two vehicle platforms are compared relative to their diesel-only baseline values over transient and steady state testing. Over a transient cycle, CO2 emissions are reduced by up to 9%; however, methane (CH4) emissions due to incomplete combustion lead to CO2e emissions that are 50-127% higher than the equivalent diesel vehicle. Oxidation catalysts evaluated on the vehicles at steady state reduced CH4 emissions by at most 15% at exhaust gas temperatures representative of transient conditions. This study highlights that control of CH4 emissions and improved control of in-cylinder CH4 combustion are required to reduce total GHG emissions of dual fuel HGVs relative to diesel vehicles.

Global civil aviation black carbon emissions.

Aircraft black carbon (BC) emissions contribute to climate forcing, but few estimates of BC emitted by aircraft at cruise exist. For the majority of aircraft engines the only BC-related measurement available is smoke number (SN)-a filter based optical method designed to measure near-ground plume visibility, not mass. While the first order approximation (FOA3) technique has been developed to estimate BC mass emissions normalized by fuel burn [EI(BC)] from SN, it is shown that it underestimates EI(BC) by >90% in 35% of directly measured cases (R(2) = -0.10). As there are no plans to measure BC emissions from all existing certified engines-which will be in service for several decades-it is necessary to estimate EI(BC) for existing aircraft on the ground and at cruise. An alternative method, called FOX, that is independent of the SN is developed to estimate BC emissions. Estimates of EI(BC) at ground level are significantly improved (R(2) = 0.68), whereas estimates at cruise are within 30% of measurements. Implementing this approach for global civil aviation estimated aircraft BC emissions are revised upward by a factor of ~3. Direct radiative forcing (RF) due to aviation BC emissions is estimated to be ~9.5 mW/m(2), equivalent to ~1/3 of the current RF due to aviation CO2 emissions.