Machine Learning Approach To Estimate Hourly Exposure to Fine Particulate Matter for Urban, Rural, and Remote Populations during Wildfire Seasons

Seasons of smoke and fire: preparing health systems for improved performance before, during, and after wildfires

Increased frequency, intensity, and duration of wildfires are intensifying exposure to direct and smoke-related hazards in many areas, leading to evacuation and smoke-related effects on health and health systems that can affect regions extending over thousands of kilometres. Effective preparation and response are currently hampered by inadequate training, continued siloing of disciplines, insufficient finance, and inadequate coordination between health systems and governance at municipal, regional, national, and international levels. This Review highlights the key health and health systems considerations before, during, and after wildfires, and outlines how a health system should respond to optimise population health outcomes now and into the future. The focus is on the implications of wildfires for air quality, mental health, and emergency management, with elements of international policy and finance also addressed. We discuss commonalities of existing climate-resilient health care and disaster management frameworks and integrate them into an approach that addresses issues of financing, leadership and governance, health workforce, health information systems, infrastructure, supply chain, technologies, community interaction and health-care delivery, before, during, and after a wildfire season. This Review is a practical briefing for leaders and health professionals facing severe wildfire seasons and a call to break down silos and join with other disciplines to proactively plan for and fund innovation and coordination in service of a healthier future.

Weather and the City: Machine Learning for Predicting and Attributing Fine Scale Air Quality to Meteorological and Urban Determinants

Urban air quality persists as a global concern, with critical health implications. This study employs a combination of machine learning (gradient boosting regression, GBR) and spatial analysis to better understand the key drivers behind air pollution and its prediction and mitigation strategies. Focusing on New York City as a representative urban area, we investigate the interplay between urban characteristics and weather factors, showing that urban features, including traffic-related parameters and urban morphology, emerge as crucial predictors for pollutants closely associated with vehicular emissions, such as elemental carbon (EC) and nitrogen oxides (NOx). Conversely, pollutants with secondary formation pathways (e.g., PM2.5) or stemming from nontraffic sources (e.g., sulfur dioxide, SO2) are predominantly influenced by meteorological conditions, particularly wind speed and maximum daily temperature. Urban characteristics are shown to act over spatial scales of 500 × 500 m2, which is thus the footprint needed to effectively capture the impact of urban form, fabric, and function. Our spatial predictive model, needing only meteorological and urban inputs, achieves promising results with mean absolute errors ranging from 8 to 32% when using full-year data. Our approach also yields good performance when applied to the temporal mapping of spatial pollutant variability. Our findings highlight the interacting roles of urban characteristics and weather conditions and can inform urban planning, design, and policy.

Exploring pollutant joint effects in disease through interpretable machine learning

Identifying the impact of pollutants on diseases is crucial. However, assessing the health risks posed by the interplay of multiple pollutants is challenging. This study introduces the concept of Pollutants Outcome Disease, integrating multidisciplinary knowledge and employing explainable artificial intelligence (AI) to explore the joint effects of industrial pollutants on diseases. Using lung cancer as a representative case study, an extreme gradient boosting predictive model that integrates meteorological, socio-economic, pollutants, and lung cancer statistical data is developed. The joint effects of industrial pollutants on lung cancer are identified and analyzed by employing the SHAP (Shapley Additive exPlanations) interpretable machine learning technique. Results reveal substantial spatial heterogeneity in emissions from CPG and ILC, highlighting pronounced nonlinear relationships among variables. The model yielded strong predictions (an R of 0.954, an RMSE of 4283, and an R2 of 0.911) and emphasized the impact of pollutant emission amounts on lung cancer responses. Diverse joint effects patterns were observed, varying in terms of patterns, regions (frequency), and the extent of antagonistic and synergistic effects among pollutants. The study provides a new perspective for exploring the joint effects of pollutants on diseases and demonstrates the potential of AI technology to assist scientific discovery.

Predicting hourly PM2.5 concentrations in wildfire-prone areas using a SpatioTemporal Transformer model

Globally, wildfires are becoming more frequent and destructive, generating a significant amount of smoke that can transport thousands of miles. Therefore, improving air pollution forecasts from wildfires is essential and informing citizens of more frequent, accurate, and interpretable updates related to localized air pollution events. This research proposes a multi-head attention-based deep learning architecture, SpatioTemporal (ST)-Transformer, to improve spatiotemporal predictions of PM_2.5 concentrations in wildfire-prone areas. The ST-Transformer model employed a sparse attention mechanism that concentrates on the most useful contextual information across spatial, temporal, and variable-wise dimensions. The model includes critical driving factors of PM_2.5 concentrations as predicting factors, including wildfire perimeter and intensity, meteorological factors, road traffic, PM_2.5, and temporal indicators from the past 24 h. The model is trained to conduct time series forecasting on PM_2.5 concentrations at EPA's air quality stations in the greater Los Angeles area. Prediction results were compared with other existing time series forecasting methods and exhibited better performance, especially in capturing abrupt changes or spikes in PM_2.5 concentrations during wildfire situations. The attention matrix learned by the proposed model enabled interpretation of the complex spatial, temporal, and variable-wise dependencies, indicating that the model can differentiate between wildfires and non-wildfires. The ST-Transformer model's accurate predictability and interpretation capacity can help effectively monitor and predict the impacts of wildfire smoke and be applicable to other complex spatiotemporal prediction problems.

The Canadian Optimized Statistical Smoke Exposure Model (CanOSSEM): A machine learning approach to estimate national daily fine particulate matter (PM2.5) exposure

Exposure to biomass smoke has been associated with a wide range of acute and chronic health outcomes. Over the past decades, the frequency and intensity of wildfires has increased in many areas, resulting in longer smoke episodes with higher concentrations of fine particulate matter (PM_2.5). There are also many communities where seasonal open burning and residential wood heating have short- and long-term impacts on ambient air quality. Understanding the acute and chronic health effects of biomass smoke exposure requires reliable estimates of PM_2.5 concentrations during the wildfire season and throughout the year, particularly in areas without regulatory air quality monitoring stations. We have developed a machine learning approach to estimate PM_2.5 across all populated regions of Canada from 2010 to 2019. The random forest machine learning model uses potential predictor variables integrated from multiple data sources and estimates daily mean (24-hour) PM_2.5 concentrations at a 5 km × 5 km spatial resolution. The training and prediction datasets were generated using observations from National Air Pollution Surveillance (NAPS) network. The Root Mean Squared Error (RMSE) between predicted and observed PM_2.5 concentrations was 2.96 μg/m³ for the entire prediction set, and more than 96 % of the predictions were within 5 μg/m³ of the NAPS PM_2.5 measurements. The model was evaluated using 10-fold, leave one-region-out, and leave-one-year-out cross-validations. Overall, CanOSSEM performed well but performance was sensitive to removal of large wildfire events such as the Fort McMurray interface fire in May 2016 or the extreme 2017 and 2018 wildfire seasons in British Columbia. Exposure estimates from CanOSSEM will be useful for epidemiologic studies on the acute and chronic health effects associated with PM_2.5 exposure, especially for populations affected by biomass smoke where routine air quality measurements are not available.

Uncertainty in Health Impact Assessments of Smoke From a Wildfire Event

Wildfires cause elevated air pollution that can be detrimental to human health. However, health impact assessments associated with emissions from wildfire events are subject to uncertainty arising from different sources. Here, we quantify and compare major uncertainties in mortality and morbidity outcomes of exposure to fine particulate matter (PM2.5) pollution estimated for a series of wildfires in the Southeastern U.S. We present an approach to compare uncertainty in estimated health impacts specifically due to two driving factors, wildfire-related smoke PM2.5 fields and variability in concentration-response parameters from epidemiologic studies of ambient and smoke PM2.5. This analysis, focused on the 2016 Southeastern wildfires, suggests that emissions from these fires had public health consequences in North Carolina. Using several methods based on publicly available monitor data and atmospheric models to represent wildfire-attributable PM2.5, we estimate impacts on several health outcomes and quantify associated uncertainty. Multiple concentration-response parameters derived from studies of ambient and wildfire-specific PM2.5 are used to assess health-related uncertainty. Results show large variability and uncertainty in wildfire impact estimates, with comparable uncertainties due to the smoke pollution fields and health response parameters for some outcomes, but substantially larger health-related uncertainty for several outcomes. Consideration of these uncertainties can support efforts to improve estimates of wildfire impacts and inform fire-related decision-making.

Design of distributed WSNs fire remote monitoring system based on fuzzy algorithm

The traditional distributed WSNs fire remote monitoring system has single monitoring variables and incomplete fire detection data, which leads to large monitoring error and long delay. A distributed WSNs fire remote monitoring system based on fuzzy algorithm is designed. The hardware part of the system consists of distributed WSNs fire remote monitor, air temperature and humidity parameters acquisition, LCD unit and system power supply unit. The remote fire monitor is designed by using microprocessor C8051F060, and the centralized monitoring of information is realized by using distributed WSNs. Based on this, the fuzzy algorithm is introduced to standardize the fire detection data, and the fuzzy similar matrix is established. According to the improved similarity coefficient, the matrix is solved, the fuzzy equivalent matrix is calculated, and the optimal threshold value of fuzzy monitoring is determined. The fuzzy language monitoring rules are set by using three fuzzy variables of current, temperature and smoke to complete the design of distributed WSNs fire remote monitoring system. The simulation results show that: compared with the traditional fire monitoring system, the system designed in this paper has higher throughput limit, shorter delay, and the accuracy rate of monitoring and alarm is higher than 95%. The experimental results show that the system has good generalization ability and is suitable for large-scale high-rise buildings and large-scale networks.

The Summer 2019-2020 Wildfires in East Coast Australia and Their Impacts on Air Quality and Health in New South Wales, Australia

The 2019-2020 summer wildfire event on the east coast of Australia was a series of major wildfires occurring from November 2019 to end of January 2020 across the states of Queensland, New South Wales (NSW), Victoria and South Australia. The wildfires were unprecedent in scope and the extensive character of the wildfires caused smoke pollutants to be transported not only to New Zealand, but also across the Pacific Ocean to South America. At the peak of the wildfires, smoke plumes were injected into the stratosphere at a height of up to 25 km and hence transported across the globe. The meteorological and air quality Weather Research and Forecasting with Chemistry (WRF-Chem) model is used together with the air quality monitoring data collected during the bushfire period and remote sensing data from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) satellites to determine the extent of the wildfires, the pollutant transport and their impacts on air quality and health of the exposed population in NSW. The results showed that the WRF-Chem model using Fire Emission Inventory (FINN) from National Center for Atmospheric Research (NCAR) to simulate the dispersion and transport of pollutants from wildfires predicted the daily concentration of PM_2.5 having the correlation (R²) and index of agreement (IOA) from 0.6 to 0.75 and 0.61 to 0.86, respectively, when compared with the ground-based data. The impact on health endpoints such as mortality and respiratory and cardiovascular diseases hospitalizations across the modelling domain was then estimated. The estimated health impact on each of the Australian Bureau of Statistics (ABS) census districts (SA4) of New South Wales was calculated based on epidemiological assumptions of the impact function and incidence rate data from the 2016 ABS and NSW Department of Health statistical health records. Summing up all SA4 census district results over NSW, we estimated that there were 247 (CI: 89, 409) premature deaths, 437 (CI: 81, 984) cardiovascular diseases hospitalizations and 1535 (CI: 493, 2087) respiratory diseases hospitalizations in NSW over the period from 1 November 2019 to 8 January 2020. The results are comparable with a previous study based only on observation data, but the results in this study provide much more spatially and temporally detailed data with regard to the health impact from the summer 2019-2020 wildfires.

Staying Ahead of the Epidemiologic Curve: Evaluation of the British Columbia Asthma Prediction System (BCAPS) During the Unprecedented 2018 Wildfire Season

Background: The modular British Columbia Asthma Prediction System (BCAPS) is designed to reduce information burden during wildfire smoke events by automatically gathering, integrating, generating, and visualizing data for public health users. The BCAPS framework comprises five flexible and geographically scalable modules: (1) historic data on fine particulate matter (PM_2.5) concentrations; (2) historic data on relevant health indicator counts; (3) PM_2.5 forecasts for the upcoming days; (4) a health forecasting model that uses the relationship between (1) and (2) to predict the impacts of (3); and (5) a reporting mechanism. Methods: The 2018 wildfire season was the most extreme in British Columbia history. Every morning BCAPS generated forecasts of salbutamol sulfate (e.g., Ventolin) inhaler dispensations for the upcoming days in 16 Health Service Delivery Areas (HSDAs) using random forest machine learning. These forecasts were compared with observations over a 63-day study period using different methods including the index of agreement (IOA), which ranges from 0 (no agreement) to 1 (perfect agreement). Some observations were compared with the same period in the milder wildfire season of 2016 for context. Results: The mean province-wide population-weighted PM_2.5 concentration over the study period was 22.0 μg/m³, compared with 4.2 μg/m³ during the milder wildfire season of 2016. The PM_2.5 forecasts underpredicted the severe smoke impacts, but the IOA was relatively strong with a population-weighted average of 0.85, ranging from 0.65 to 0.95 among the HSDAs. Inhaler dispensations increased by 30% over 2016 values. Forecasted dispensations were within 20% of the observed value in 71% of cases, and the IOA was strong with a population-weighted average of 0.95, ranging from 0.92 to 0.98. All measures of agreement were correlated with HSDA population, where BCAPS performance was better in the larger populations with more moderate smoke impacts. The accuracy of the health forecasts was partially dependent on the accuracy of the PM_2.5 forecasts, but they were robust to over- and underpredictions of PM_2.5 exposure. Conclusions: Daily reports from the BCAPS framework provided timely and reasonable insight into the population health impacts of predicted smoke exposures, though more work is necessary to improve the PM_2.5 and health indicator forecasts.

Estimation of PM2.5 Concentrations in New York State: Understanding the Influence of Vertical Mixing on Surface PM2.5 Using Machine Learning

In New York State (NYS), episodic high fine particulate matter (PM2.5) concentrations associated with aerosols originated from the Midwest, Mid-Atlantic, and Pacific Northwest states have been reported. In this study, machine learning techniques, including multiple linear regression (MLR) and artificial neural network (ANN), were used to estimate surface PM2.5 mass concentrations at air quality monitoring sites in NYS during the summers of 2016–2019. Various predictors were considered, including meteorological, aerosol, and geographic predictors. Vertical predictors, designed as the indicators of vertical mixing and aloft aerosols, were also applied. Overall, the ANN models performed better than the MLR models, and the application of vertical predictors generally improved the accuracy of PM2.5 estimation of the ANN models. The leave-one-out cross-validation results showed significant cross-site variations and were able to present the different predictor-PM2.5 correlations at the sites with different PM2.5 characteristics. In addition, a joint analysis of regression coefficients from the MLR model and variable importance from the ANN model provided insights into the contributions of selected predictors to PM2.5 concentrations. The improvements in model performance due to aloft aerosols were relatively minor, probably due to the limited cases of aloft aerosols in current datasets.

Differences in the Estimation of Wildfire-Associated Air Pollution by Satellite Mapping of Smoke Plumes and Ground-Level Monitoring

Wildfires, which are becoming more frequent and intense in many countries, pose serious threats to human health. To determine health impacts and provide public health messaging, satellite-based smoke plume data are sometimes used as a proxy for directly measured particulate matter levels. We collected data on particulate matter <2.5 μm in diameter (PM_2.5) concentration from 16 ground-level monitoring stations in the San Francisco Bay Area and smoke plume density from satellite imagery for the 2017–2018 California wildfire seasons. We tested for trends and calculated bootstrapped differences in the median PM_2.5 concentrations by plume density category on a 0–3 scale. The median PM_2.5 concentrations for categories 0, 1, 2, and 3 were 16, 22, 25, and 63 μg/m³, respectively, and there was much variability in PM_2.5 concentrations within each category. A case study of the Camp Fire illustrates that in San Francisco, PM_2.5 concentrations reached their maximum many days after the peak for plume density scores. We found that air pollution characterization by satellite imagery did not precisely align with ground-level PM_2.5 concentrations. Public health practitioners should recognize the need to combine multiple sources of data regarding smoke patterns when developing public guidance to limit the health effects of wildfire smoke.

Spatio-temporal air pollution modelling using a compositional approach

Air pollutant data are compositional in character because they describe quantitatively the parts of a whole (atmospheric composition). However, it is common to use air pollutant concentrations in statistical models without considering this characteristic of the data and, therefore, without control of common statistical problems, such as spurious correlations and subcompositional incoherence. This paper now proposes a daily multivariate spatio-temporal model with a compositional approach. The air pollution spatio-temporal model is based on a dynamic linear modelling framework with Bayesian inference. The novel modelling methodology was applied in an urban area for carbon monoxide (CO, mg·m-3), sulfur dioxide (SO2, μg·m-3), ozone (O3, μg·m-3), nitrogen dioxide (NO2, μg·m-3), and particulate matter less than 2.5 μm in aerodynamic diameter (PM2.5, μg·m-3). The proposal complemented and improved the conventional approach in air pollution modelling. The main improvements come from a fast multivariate data description, high spatial-correlation, and adequate modelling of air pollutants with high variability.

Sub-Daily Exposure to Fine Particulate Matter and Ambulance Dispatches during Wildfire Seasons: A Case-Crossover Study in British Columbia, Canada

BACKGROUND

Exposure to fine particulate matter (PM2.5) during wildfire seasons has been associated with adverse health outcomes. Previous studies have focused on daily exposure, but PM2.5 levels in smoke events can vary considerably within 1 d.

OBJECTIVES

We aimed to assess the immediate and lagged relationship between sub-daily exposure to PM2.5 and acute health outcomes during wildfire seasons in British Columbia.

METHODS

We used a time-stratified case-crossover study design to evaluate the association between modeled hourly PM2.5 and ambulance dispatches during wildfire seasons from 2010 to 2015. Distributed lag nonlinear models were used to estimate the lag-specific and cumulative odds ratios (ORs) at lags from 1 to 48 h. We examined the relationship for all dispatches and dispatches related to respiratory, circulatory, and diabetic conditions, identified by codes for ambulance dispatch (AD), paramedic assessment (PA) or hospital diagnosis (HD).

RESULTS

Increased respiratory health outcomes were observed within 1 h of exposure to a 10-μg/m3 increase in PM2.5. The 48-h cumulative OR [95% confidence interval (CI)] was 1.038 (1.009, 1.067) for the AD code Breathing Problems and 1.098 (1.013, 1.189) for PA code Asthma/COPD. The point estimates were elevated within 1 h for the PA code for Myocardial Infarction and HD codes for Ischemic Heart Disease, which had 24-h cumulative ORs of 1.104 (0.915, 1.331) and 1.069 (0.983, 1.162), respectively. The odds of Diabetic AD and PA codes increased over time to a cumulative 24-h OR of 1.075 (1.001, 1.153) and 1.104 (1.015, 1.202) respectively.

CONCLUSIONS

We found increased PM2.5 during wildfire seasons was associated with some respiratory and cardiovascular outcomes within 1 h following exposure, and its association with diabetic outcomes increased over time. Cumulative effects were consistent with those reported elsewhere in the literature. These results warrant further investigation and may have implications for the appropriate time scale of public health actions. https://doi.org/10.1289/EHP5792.

Environmental Particulate Matter Levels during 2017 Large Forest Fires and Megafires in the Center Region of Portugal: A Public Health Concern?

This work characterizes the dimension and the exceptionality of 2017 large- and mega-fires that occurred in the center region of Portugal through the assessment of their impact on the ambient levels of particulate matter (PM10 and PM2.5), retrieved from local monitoring stations, and the associated public health risks. PM10 and PM2.5 concentrations were increased during the occurrence of large fires and megafires, with daily concentrations exceeding the European/national guidelines in 7-14 and 1-12 days of 2017 (up to 704 µg/m3 for PM10 and 46 µg/m3 for PM2.5), respectively. PM10 concentrations were correlated with total burned area (0.500 < r < 0.949; p > 0.05) and with monthly total burned area/distance2 (0.500 < r < 0.667; p > 0.05). The forest fires of 2017 took the life of 112 citizens. A total of 474 cases of hospital admissions due to cardiovascular diseases and 3524 cases of asthma incidence symptoms per 100,000 individuals at risk were assessed due to exposure to 2017 forest fires. Real-time and in situ PM methodologies should be combined with protection action plans to reduce public health risks. Portuguese rural stations should monitor other health-relevant pollutants (e.g., carbon monoxide and volatile organic compounds) released from wildfires to allow performing more robust and comprehensive measurements that will allow a better assessment of the potential health risks for the exposed populations.