Ozone response modeling to NOx and VOC emissions: Examining machine learning models.

Current machine learning (ML) applications in atmospheric science focus on forecasting and bias correction for numerical modeling estimations, but few studies examined the nonlinear response of their predictions to precursor emissions. This study uses ground-level maximum daily 8-hour ozone average (MDA8 O3) as an example to examine O3 responses to local anthropogenic NOx and VOC emissions in Taiwan by Response Surface Modeling (RSM). Three different datasets for RSM were examined, including the Community Multiscale Air Quality (CMAQ) model data, ML-measurement-model fusion (ML-MMF) data, and ML data, which respectively represent direct numerical model predictions, numerical predictions adjusted by observations and other auxiliary data, and ML predictions based on observations and other auxiliary data. The results show that both ML-MMF (r = 0.93-0.94) and ML predictions (r = 0.89-0.94) present significantly improved performance in the benchmark case compared with CMAQ predictions (r = 0.41-0.80). While ML-MMF isopleths exhibit O3 nonlinearity close to actual responses due to their numerical base and observation-based correction, ML isopleths present biased predictions concerning their different controlled ranges of O3 and distorted O3 responses to NOx and VOC emission ratios compared with ML-MMF isopleths, which implies that using data without support from CMAQ modeling to predict the air quality could mislead the controlled targets and future trends. Meanwhile, the observation-corrected ML-MMF isopleths also emphasize the impact of transboundary pollution from mainland China on the regional O3 sensitivity to local NOx and VOC emissions, which transboundary NOx would make all air quality regions in April more sensitive to local VOC emissions and limit the potential effort by reducing local emissions. Future ML applications in atmospheric science like forecasting or bias correction should provide interpretability and explainability, except for meeting statistical performance and providing variable importance. Assessment with interpretable physical and chemical mechanisms and constructing a statistically robust ML model should be equally important.

Localized energy burden, concentrated disadvantage, and the feminization of energy poverty.

Energy burden directly influences households' health and safety. Amid a growing literature on energy, poverty and gender remains relatively understudied. We evaluate socioeconomic, geographic, and health factors as multidimensions of concentrated disadvantage that magnify energy burden in the United States over time. We show that the energy burden is more pronounced in disadvantaged counties with larger elderly, impoverished, disabled people, and racialized populations where people do not have health insurance. Neighborhoods with households headed by women of color (especially Black women) are more likely to face a high energy burden, which worsened during the COVID-19 pandemic. Although energy costs are often regarded as an individual responsibility, these findings illustrate the feminization of energy poverty and indicate the need for an intersectional and interdisciplinary framework in devising energy policy directed to households with the most severe energy burden.

How limitations in energy access, poverty, and socioeconomic disparities compromise health interventions for outbreaks in urban settings.

Low-income households (LIHs) have experienced increased poverty and inaccess to healthcare services during the COVID-19 pandemic, limiting their ability to adhere to health-protective behaviors. We use an epidemiological model to show how a households' inability to adopt social distancing, owing to constraints in utility and healthcare expenditure, can drastically impact the course of disease outbreaks in five urban U.S. counties. LIHs suffer greater burdens of disease and death than higher income households, while functioning as a consistent source of virus exposure for the entire community due to socioeconomic barriers to following public health guidelines. These impacts worsened when social distancing policy could not be imposed. Health interventions combining social distancing and LIH resource protection strategies (e.g., utility and healthcare access) were the most effective in limiting virus spread for all income levels. Policies need to address the multidimensionality of energy, housing, and healthcare access for future disaster management.

Quantifying spatial heterogeneity of vulnerability to short-term PM2.5 exposure with data fusion framework.

The current estimations of the burden of disease (BD) of PM2.5 exposure is still potentially biased by two factors: ignorance of heterogeneous vulnerabilities at diverse urbanization levels and reliance on the risk estimates from existing literature, usually from different locations. Our objectives are (1) to build up a data fusion framework to estimate the burden of PM2.5 exposure while evaluating local risks simultaneously and (2) to quantify their spatial heterogeneity, relationship to land-use characteristics, and derived uncertainties when calculating the disease burdens. The feature of this study is applying six local databases to extract PM2.5 exposure risk and the BD information, including the risks of death, cardiovascular disease (CVD), and respiratory disease (RD), and their spatial heterogeneities through our data fusion framework. We applied the developed framework to Tainan City in Taiwan as a use case estimated the risks by using 2006-2016 emergency department visit data, air quality monitoring data, and land-use characteristics and further estimated the BD caused by daily PM2.5 exposure in 2013. Our results found that the risks of CVD and RD in highly urbanized areas and death in rural areas could reach 1.20-1.57 times higher than average. Furthermore, we performed a sensitivity analysis to assess the uncertainty of BD estimations from utilizing different data sources, and the results showed that the uncertainty of the BD estimations could be contributed by different PM2.5 exposure data (20-32%) and risk values (0-86%), especially for highly urbanized areas. In conclusion, our approach for estimating BD based on local databases has the potential to be generalized to the developing and overpopulated countries and to support local air quality and health management plans.

Evaluating the impact of mobility on COVID-19 pandemic with machine learning hybrid predictions.

COVID-19 pandemic had expanded to the US since early 2020 and has caused nationwide economic loss and public health crisis. Until now, although the US has the most confirmed cases in the world and are still experiencing an increasing pandemic, several states insisted to re-open business activities and colleges while announced strict control measures. To provide a quantitative reference for official strategies, predicting the near future trend based on finer spatial resolution data and presumed scenarios are urgently needed. In this study, the first attempted COVID-19 case predicting model based on county-level demographic, environmental, and mobility data was constructed with multiple machine learning techniques and a hybrid framework. Different scenarios were also applied to selected metropolitan counties including New York City, Cook County in Illinois, Los Angeles County in California, and Miami-Dade County in Florida to assess the impact from lockdown, Phase I, and Phase III re-opening. Our results showed that, for selected counties, the mobility decreased substantially after the lockdown but kept increasing with an apparent weekly pattern, and the weekly pattern of mobility and infections implied high infections during the weekend. Meanwhile, our model was successfully built up, and the scenario assessment results indicated that, compared with Phase I re-opening, a 1-week and a 2-week lockdown could reduce 4%-29% and 15%-55% infections, respectively, in the future week, while 2-week Phase III re-opening could increase 16%-80% infections. We concluded that the mandatory orders in metropolitan counties such lockdown should last longer than one week, the effect could be observed. The impact of lockdown or re-opening was also county-dependent and varied with the local pandemic. In future works, we expect to involve a longer period of data, consider more county-dependent factors, and employ more sophisticated techniques to decrease the modeling uncertainty and apply it to counties nationally and other countries.

Transient risk of ambient fine particulate matter on hourly cardiovascular events in Tainan City, Taiwan.

BACKGROUND: The association between daily changes in ambient fine particulate matter (PM2.5) and cardiovascular diseases have been well established in mechanistic, epidemiologic and exposure studies. Only a few studies examined the effect of hourly variations in air pollution on triggering cardiovascular events. Whether the current PM2.5 standards can protect vulnerable individuals with chronic cardiovascular diseases remain uncertain. METHODS: we conducted a time-stratified, case-crossover study to assess the associations between hourly changes in PM2.5 levels and the vascular disease onset in residents of Tainan City, Taiwan, visiting Emergency Room of Chi Mei Medical Center between January 2006 and December 2016. There were 26,749 cases including 10,310 females (38.5%) and 16,439 males (61.5%) identified. The time of emergency visit was identified as the onset for each case and control cases were selected as the same times on other days, on the same day of the week in the same month and year respectively. Residential address was used to identify the ambient air pollution exposure concentrations from the closest station. Conditional logistic regression with the stepwise selection method was used to estimate adjusted odds ratios (ORs) for the association. RESULTS: When we only included cases occurring at PM2.5>10 µg/m3 and PM2.5>25 µg/m3, very significant ORs could be observed for 10 µg/m3 increases in PM2.5 at 0 and 1 hour, implying fine particulate exposure could promptly trigger vascular disease events. Moreover, a very clear increase in risk could be observed with cumulative exposure from 0 to 48 hours, especially in those cases where PM2.5>25 µg/m3. CONCLUSIONS: Our study demonstrated that transient and low concentrations of ambient PM2.5 trigger adult vascular disease events, especially cerebrovascular disease, regardless of age, sex, and exposure timing. Warning and delivery systems should be setup to protect people from these prompt adverse health impacts.

Source Apportionment of Volatile Organic Compounds (VOCs) by Positive Matrix Factorization (PMF) supported by Model Simulation and Source Markers - Using Petrochemical Emissions as a Showcase.

This study demonstrates the use of positive matrix factorization (PMF) in a region with a major Petrochemical Complex, a prominent source of volatile organic compounds (VOCs), as a showcase of PMF applications. The PMF analysis fully exploited the quality and quantity of the observation data, sufficed by a cluster of 9 monitoring sites within a 20 km radius of the petro-complex. Each site provided continuous data of 54 speciated VOCs and meteorological variables. Wind characteristics were highly seasonal and played a decisive role in the source-receptor relationship, hence the dataset was divided into three sub-sets in accordance with the prevailing wind flows. A full year of real-time data were analyzed by PMF to resolve into various distinct source types including petrochemical, urban, evaporative, long-range air parcels, etc., with some sites receiving more petro-influence than others. To minimize subjectivity in the assignment of the PMF source factors, as commonly seen in some PMF works, this study attempted to solidify PMF results by supporting with two tools of spatially/temporally resolved air-quality model simulations and observation data. By exploiting the two supporting tools, the dynamic process of individual sources to a receptor were rationalized. Percent contributions from these sources to the receptor sites were calculated by summing over the occurrence of different source types. Interestingly, although the Petro-complex is the single largest local VOC source in the 20 km radius study domain, all monitoring sites in the region received far less influence from the Petro-complex than from other emission types within or outside the region, which together add up to more than 70% of the total VOC abundance.

Source apportionment of particulate matter and selected volatile organic compounds with multiple time resolution data.

Fine particulate matter (PM2.5) and volatile organic compounds (VOCs) co-exist in ambient air and contribute to adverse health effects in human populations. This study was conducted to demonstrate the feasibility of utilizing a composite data set which included PM2.5 and VOC data with multiple time resolutions for source apportionment. Hourly VOC and 12-h PM2.5 speciation data were combined into an improved source apportionment model to quantify different pollutant source contributions to PM2.5 and VOC mixtures. Five factors were retrieved, including vehicle 1, vehicle 2, industrial processing, transported regional, and secondary pollution sources. The largest contributors were vehicular emissions for VOCs (62%) and PM2.5 (35%). Nonetheless, transported regional and secondary pollution sources accounted for a noteworthy portion of PM2.5 (27% and 25%, respectively) relative to VOCs (8% and 5%, respectively). Additional sensitivity analyses showed that excluding the PM2.5 data or reducing the associated temporal resolution (12-h to 24-h) retrieved fewer source factors and increased the errors of source contribution estimates.